Daily Papers

Motif-scaffolding systems commonly follow a generate-then-filter paradigm, in which candidate proteins are generated independently and structural evaluation is used primarily for terminal screening or ranking. This paradigm underuses evaluation: failed predictions contain state-specific evidence about whether a design requires repair of motif geometry, global foldability, or other structural constraints. We introduce ELMS (Evidence-based LLM-guided Monte Carlo Search), an evaluator-in-the-loop search framework for motif scaffolding that turns such evaluator feedback into targeted design actions. Effective reuse of structural feedback is nontrivial because different scaffold states exhibit different failure modes, and repeatedly refining a single trajectory can prematurely commit computation to an unproductive region of sequence space. ELMS therefore retains evaluated scaffolds as persistent search states: a Critic Agent diagnoses state-local structural failures, a Policy Agent selects targeted operators with execution parameters, motif-locked operators realize legal sequence modifications, and MCTS determines which historical states should receive further design effort. Under the standard GeomMotif protocol (100 candidates per task), ELMS achieves Successful rates of 86.41\% on single-motif tasks and 84.57\% on paired-motif tasks, exceeding the strongest prior baseline by 19.3 and 21.9 percentage points, respectively. On MotifBench, under a matched 100-candidate search budget, it solves 26.7 of 30 tasks on average (88.89\% Task Success), compared with 16.0 tasks (53.33\%) for the strongest baseline. These results establish ELMS as an effective approach for converting structural evaluation from a terminal filter into actionable guidance for iterative motif scaffolding.

Self-evaluation using large language models (LLMs) has proven valuable not only in benchmarking but also methods like reward modeling, constitutional AI, and self-refinement. But new biases are introduced due to the same LLM acting as both the evaluator and the evaluatee. One such bias is self-preference, where an LLM evaluator scores its own outputs higher than others' while human annotators consider them of equal quality. But do LLMs actually recognize their own outputs when they give those texts higher scores, or is it just a coincidence? In this paper, we investigate if self-recognition capability contributes to self-preference. We discover that, out of the box, LLMs such as GPT-4 and Llama 2 have non-trivial accuracy at distinguishing themselves from other LLMs and humans. By fine-tuning LLMs, we discover a linear correlation between self-recognition capability and the strength of self-preference bias; using controlled experiments, we show that the causal explanation resists straightforward confounders. We discuss how self-recognition can interfere with unbiased evaluations and AI safety more generally.

Selecting a single high-quality output from multiple stochastic generations remains a fundamental challenge for large language models (LLMs), particularly in open-ended tasks where no canonical answer exists. While Best-of-N and self-consistency methods show that aggregating multiple generations can improve performance, existing approaches typically rely on external evaluators, reward models, or exact string-match voting, limiting their applicability and efficiency. We propose Mode Extraction (ModeX), an evaluator-free Best-of-N selection framework that generalizes majority voting to open-ended text generation by identifying the modal output representing the dominant semantic consensus among generated texts. ModeX constructs a similarity graph over candidate generations and recursively applies spectral clustering to select a representative centroid, without requiring additional inference or auxiliary models. We further instantiate this selection principle as ModeX-Lite, an improved version of ModeX with early pruning for efficiency. Across open-ended tasks -- including text summarization, code generation, and mathematical reasoning -- our approaches consistently outperform standard single- and multi-path baselines, providing a computationally efficient solution for robust open-ended text generation. Code is released in https://github.com/deeplearning-wisc/ModeX.

LLM-based evaluators have rapidly become a standard tool for machine translation evaluation, yet the design choices that shape their behavior remain poorly understood. We present a systematic study of how instruction design, few-shot demonstrations, backbone model choice, and multi-step workflows affect span-level and ranking-based MT evaluation. We collect new expert RATE annotations on WMT24 and WMT25 English-Russian translations, with transfer validation on MQM- and ESA-annotated benchmarks across five additional language pairs. Our experiments reveal four robust findings: instruction structure shapes evaluator behavior, high-quality demonstrations matter more than quantity, frontier backbones exhibit distinct evaluator profiles (recall-, precision-, or ranking-oriented), and fusing outputs across diverse backbones outperforms single-evaluator multi-call setups. Although evaluator design exposes a large combinatorial space, several discovered design principles transfer consistently across annotation protocols and language pairs. We synthesize these findings into ORBIT, a family of LLM-based MT evaluators. Our strongest configurations achieve expert-level performance in error span detection and in Accuracy and Fluency ranking at roughly 20x lower cost than expert human annotation.

Recent advancements in large language models (LLMs) on language modeling and emergent capabilities make them a promising reference-free evaluator of natural language generation quality, and a competent alternative to human evaluation. However, hindered by the closed-source or high computational demand to host and tune, there is a lack of practice to further calibrate an off-the-shelf LLM-based evaluator towards better human alignment. In this work, we propose AutoCalibrate, a multi-stage, gradient-free approach to automatically calibrate and align an LLM-based evaluator toward human preference. Instead of explicitly modeling human preferences, we first implicitly encompass them within a set of human labels. Then, an initial set of scoring criteria is drafted by the language model itself, leveraging in-context learning on different few-shot examples. To further calibrate this set of criteria, we select the best performers and re-draft them with self-refinement. Our experiments on multiple text quality evaluation datasets illustrate a significant improvement in correlation with expert evaluation through calibration. Our comprehensive qualitative analysis conveys insightful intuitions and observations on the essence of effective scoring criteria.

Large Language Models (LLMs) are increasingly relied upon to evaluate text outputs of other LLMs, thereby influencing leaderboards and development decisions. However, concerns persist over the accuracy of these assessments and the potential for misleading conclusions. In this work, we investigate the effectiveness of LLMs as evaluators for text generation tasks. We propose FBI, a novel framework designed to examine the proficiency of Evaluator LLMs in assessing four critical abilities in other LLMs: factual accuracy, instruction following, coherence in long-form writing, and reasoning proficiency. By introducing targeted perturbations in answers generated by LLMs, that clearly impact one of these key capabilities, we test whether an Evaluator LLM can detect these quality drops. By creating a total of 2400 perturbed answers covering 22 perturbation categories, we conduct a comprehensive study using different evaluation strategies on five prominent LLMs commonly used as evaluators in the literature. Our findings reveal significant shortcomings in current Evaluator LLMs, which failed to identify quality drops in over 50\% of cases on average. Single-answer and pairwise evaluations demonstrated notable limitations, whereas reference-based evaluations showed comparatively better performance. These results underscore the unreliable nature of current Evaluator LLMs and advocate for cautious implementation in practical applications. Code and data are available at https://github.com/AI4Bharat/FBI.

This research investigates prompt designs of evaluating generated texts using large language models (LLMs). While LLMs are increasingly used for scoring various inputs, creating effective prompts for open-ended text evaluation remains challenging due to model sensitivity and subjectivity in evaluation of text generation. Our study experimented with different prompt structures, altering the sequence of output instructions and including explanatory reasons. We found that the order of presenting reasons and scores significantly influences LLMs' scoring, with a different level of rule understanding in the prompt. An additional optimization may enhance scoring alignment if sufficient data is available. This insight is crucial for improving the accuracy and consistency of LLM-based evaluations.

Model-based evaluation is at the heart of successful model development -- as a reward model for training, and as a replacement for human evaluation. To train such evaluators, the standard approach is to collect a large amount of human preference judgments over model responses, which is costly and the data becomes stale as models improve. In this work, we present an approach that aims to im-prove evaluators without human annotations, using synthetic training data only. Starting from unlabeled instructions, our iterative self-improvement scheme generates contrasting model outputs and trains an LLM-as-a-Judge to produce reasoning traces and final judgments, repeating this training at each new iteration using the improved predictions. Without any labeled preference data, our Self-Taught Evaluator can improve a strong LLM (Llama3-70B-Instruct) from 75.4 to 88.3 (88.7 with majority vote) on RewardBench. This outperforms commonly used LLM judges such as GPT-4 and matches the performance of the top-performing reward models trained with labeled examples.

The vision of a broadly capable and goal-directed agent, such as an Internet-browsing agent in the digital world and a household humanoid in the physical world, has rapidly advanced, thanks to the generalization capability of foundation models. Such a generalist agent needs to have a large and diverse skill repertoire, such as finding directions between two travel locations and buying specific items from the Internet. If each skill needs to be specified manually through a fixed set of human-annotated instructions, the agent's skill repertoire will necessarily be limited due to the quantity and diversity of human-annotated instructions. In this work, we address this challenge by proposing Proposer-Agent-Evaluator, an effective learning system that enables foundation model agents to autonomously discover and practice skills in the wild. At the heart of PAE is a context-aware task proposer that autonomously proposes tasks for the agent to practice with context information of the environment such as user demos or even just the name of the website itself for Internet-browsing agents. Then, the agent policy attempts those tasks with thoughts and actual grounded operations in the real world with resulting trajectories evaluated by an autonomous VLM-based success evaluator. The success evaluation serves as the reward signal for the agent to refine its policies through RL. We validate PAE on challenging vision-based web navigation, using both real-world and self-hosted websites from WebVoyager and WebArena.To the best of our knowledge, this work represents the first effective learning system to apply autonomous task proposal with RL for agents that generalizes real-world human-annotated benchmarks with SOTA performances. Our open-source checkpoints and code can be found in https://yanqval.github.io/PAE/

Although applications involving long-context inputs are crucial for the effective utilization of large language models (LLMs), they also result in increased computational costs and reduced performance. To address this challenge, we propose an efficient, training-free prompt compression method that retains key information within compressed prompts. We identify specific attention heads in transformer-based LLMs, which we designate as evaluator heads, that are capable of selecting tokens in long inputs that are most significant for inference. Building on this discovery, we develop EHPC, an Evaluator Head-based Prompt Compression method, which enables LLMs to rapidly "skim through" input prompts by leveraging only the first few layers with evaluator heads during the pre-filling stage, subsequently passing only the important tokens to the model for inference. EHPC achieves state-of-the-art results across two mainstream benchmarks: prompt compression and long-context inference acceleration. Consequently, it effectively reduces the complexity and costs associated with commercial API calls. We further demonstrate that EHPC attains competitive results compared to key-value cache-based acceleration methods, thereby highlighting its potential to enhance the efficiency of LLMs for long-context tasks.

With the rapid growth of video generative models (VGMs), it is essential to develop reliable and comprehensive automatic metrics for AI-generated videos (AIGVs). Existing methods either use off-the-shelf models optimized for other tasks or rely on human assessment data to train specialized evaluators. These approaches are constrained to specific evaluation aspects and are difficult to scale with the increasing demands for finer-grained and more comprehensive evaluations. To address this issue, this work investigates the feasibility of using multimodal large language models (MLLMs) as a unified evaluator for AIGVs, leveraging their strong visual perception and language understanding capabilities. To evaluate the performance of automatic metrics in unified AIGV evaluation, we introduce a benchmark called UVE-Bench. UVE-Bench collects videos generated by state-of-the-art VGMs and provides pairwise human preference annotations across 15 evaluation aspects. Using UVE-Bench, we extensively evaluate 16 MLLMs. Our empirical results suggest that while advanced MLLMs (e.g., Qwen2VL-72B and InternVL2.5-78B) still lag behind human evaluators, they demonstrate promising ability in unified AIGV evaluation, significantly surpassing existing specialized evaluation methods. Additionally, we conduct an in-depth analysis of key design choices that impact the performance of MLLM-driven evaluators, offering valuable insights for future research on AIGV evaluation. The code is available at https://github.com/bytedance/UVE.

Scientific innovation is undergoing a paradigm shift driven by the rapid advancement of Large Language Models (LLMs). As science faces mounting challenges including information overload, disciplinary silos, and diminishing returns on conventional research methods, LLMs are emerging as powerful agents capable not only of enhancing scientific workflows but also of participating in and potentially leading the innovation process. Existing surveys mainly focus on different perspectives, phrases, and tasks in scientific research and discovery, while they have limitations in understanding the transformative potential and role differentiation of LLM. This survey proposes a comprehensive framework to categorize the evolving roles of LLMs in scientific innovation across three hierarchical levels: Evaluator, Collaborator, and Scientist. We distinguish between LLMs' contributions to structured scientific research processes and open-ended scientific discovery, thereby offering a unified taxonomy that clarifies capability boundaries, evaluation criteria, and human-AI interaction patterns at each level. Through an extensive analysis of current methodologies, benchmarks, systems, and evaluation metrics, this survey delivers an in-depth and systematic synthesis on LLM-driven scientific innovation. We present LLMs not only as tools for automating existing processes, but also as catalysts capable of reshaping the epistemological foundations of science itself. This survey offers conceptual clarity, practical guidance, and theoretical foundations for future research, while also highlighting open challenges and ethical considerations in the pursuit of increasingly autonomous AI-driven science. Resources related to this survey can be accessed on GitHub at: https://github.com/haoxuan-unt2024/llm4innovation.

Large Language Models (LLMs) are increasingly employed as automated evaluators to assess the safety of generated content, yet their reliability in this role remains uncertain. This study evaluates a diverse set of 11 LLM judge models across critical safety domains, examining three key aspects: self-consistency in repeated judging tasks, alignment with human judgments, and susceptibility to input artifacts such as apologetic or verbose phrasing. Our findings reveal that biases in LLM judges can significantly distort the final verdict on which content source is safer, undermining the validity of comparative evaluations. Notably, apologetic language artifacts alone can skew evaluator preferences by up to 98\%. Contrary to expectations, larger models do not consistently exhibit greater robustness, while smaller models sometimes show higher resistance to specific artifacts. To mitigate LLM evaluator robustness issues, we investigate jury-based evaluations aggregating decisions from multiple models. Although this approach both improves robustness and enhances alignment to human judgements, artifact sensitivity persists even with the best jury configurations. These results highlight the urgent need for diversified, artifact-resistant methodologies to ensure reliable safety assessments.

Finetuning specialized generative evaluators has emerged as a popular paradigm to meet the increasing demand for scalable evaluation during both training and test-time. However, recent work has largely focused on applying new methodology, such as reinforcement learning (RL), to training evaluators, shying away from large-scale, data-driven development. In this work, we focus on data scaling, curating a set of 2.5M samples spanning five unique evaluation tasks (pairwise, step-level, reference-free and reference-based verification, and single rating) and multiple domains focused on reasoning evaluation. With our data, we train Foundational Automatic Reasoning Evaluators (FARE), a family of 8B and 20B (with 3.6B active) parameter evaluators, with a simple iterative rejection-sampling supervised finetuning (SFT) approach. FARE-8B challenges larger specialized RL-trained evaluators and FARE-20B sets the new standard for open-source evaluators, surpassing specialized 70B+ evaluators. Beyond static benchmarks, we evaluate FARE in real-world tasks: As inference-time rerankers, FARE-20B achieves near-oracle performance on MATH. As verifiers in RL training, FARE improves the downstream RL-trained model performance by up to 14.1% vs. string-matching verifiers. When initialized from FARE, a continually-finetuned FARE-Code outperforms gpt-oss-20B by 65% on evaluating test-case quality.

Video matting remains limited by the scale and realism of existing datasets. While leveraging segmentation data can enhance semantic stability, the lack of effective boundary supervision often leads to segmentation-like mattes lacking fine details. To this end, we introduce a learned Matting Quality Evaluator (MQE) that assesses semantic and boundary quality of alpha mattes without ground truth. It produces a pixel-wise evaluation map that identifies reliable and erroneous regions, enabling fine-grained quality assessment. The MQE scales up video matting in two ways: (1) as an online matting-quality feedback during training to suppress erroneous regions, providing comprehensive supervision, and (2) as an offline selection module for data curation, improving annotation quality by combining the strengths of leading video and image matting models. This process allows us to build a large-scale real-world video matting dataset, VMReal, containing 28K clips and 2.4M frames. To handle large appearance variations in long videos, we introduce a reference-frame training strategy that incorporates long-range frames beyond the local window for effective training. Our MatAnyone 2 achieves state-of-the-art performance on both synthetic and real-world benchmarks, surpassing prior methods across all metrics.

Large Language Models (LLMs) often exhibit sycophancy, distorting responses to align with user beliefs, notably by readily agreeing with user counterarguments. Paradoxically, LLMs are increasingly adopted as successful evaluative agents for tasks such as grading and adjudicating claims. This research investigates that tension: why do LLMs show sycophancy when challenged in subsequent conversational turns, yet perform well when evaluating conflicting arguments presented simultaneously? We empirically tested these contrasting scenarios by varying key interaction patterns. We find that state-of-the-art models: (1) are more likely to endorse a user's counterargument when framed as a follow-up from a user, rather than when both responses are presented simultaneously for evaluation; (2) show increased susceptibility to persuasion when the user's rebuttal includes detailed reasoning, even when the conclusion of the reasoning is incorrect; and (3) are more readily swayed by casually phrased feedback than by formal critiques, even when the casual input lacks justification. Our results highlight the risk of relying on LLMs for judgment tasks without accounting for conversational framing.

Large Vision-Language Models (VLMs) are increasingly used to evaluate outputs of other models, for image-to-text (I2T) tasks such as visual question answering, and text-to-image (T2I) generation tasks. Despite this growing reliance, the reliability of these Evaluator VLMs remains under explored. In this work, we systematically evaluate the reliability of Evaluator VLMs across both I2T and T2I tasks. We introduce targeted perturbations that degrade output quality along key error dimensions, including object hallucinations, spatial reasoning, factual grounding, and visual fidelity. These perturbations test whether Evaluator VLMs can reliably account for these quality degrading errors in their evaluations. Using a comprehensive benchmark of over 4000 perturbed instances spanning 40 perturbation dimensions, we evaluate 4 prominent VLMs using single-answer scoring, pairwise comparison, and reference-guided paradigms. Our findings reveal that current VLM evaluators exhibit substantial blind spots: they often fail to detect perturbed outputs - in some cases exceeding 50%, struggle particularly with fine-grained compositional and spatial errors, and are often insensitive to hallucinated content that contradicts the input image. Pairwise comparison proves more reliable, though failure rates persist. These results highlight the unreliable nature of current Evaluator VLMs and urge caution in their deployment for benchmarking and development decisions. Code and data have been made publicly available.

Large Language Models (LLMs) have demonstrated impressive performance on Natural Language Processing (NLP) tasks, such as Question Answering, Summarization, and Classification. The use of LLMs as evaluators, that can rank or score the output of other models (usually LLMs) has become increasingly popular, due to the limitations of current evaluation techniques including the lack of appropriate benchmarks, metrics, cost, and access to human annotators. While LLMs are capable of handling approximately 100 languages, the majority of languages beyond the top 20 lack systematic evaluation across various tasks, metrics, and benchmarks. This creates an urgent need to scale up multilingual evaluation to ensure a precise understanding of LLM performance across diverse languages. LLM-based evaluators seem like the perfect solution to this problem, as they do not require human annotators, human-created references, or benchmarks and can theoretically be used to evaluate any language covered by the LLM. In this paper, we investigate whether LLM-based evaluators can help scale up multilingual evaluation. Specifically, we calibrate LLM-based evaluation against 20k human judgments of five metrics across three text-generation tasks in eight languages. Our findings indicate that LLM-based evaluators may exhibit bias towards higher scores and should be used with caution and should always be calibrated with a dataset of native speaker judgments, particularly in low-resource and non-Latin script languages.

Evaluating Large Language Models (LLMs) is a complex task, especially considering the intricacies of natural language understanding and the expectations for high-level reasoning. Traditional evaluations typically lean on human-based, model-based, or automatic-metrics-based paradigms, each with its own advantages and shortcomings. We introduce "Fusion-Eval", a system that employs LLMs not solely for direct evaluations, but to skillfully integrate insights from diverse evaluators. This gives Fusion-Eval flexibility, enabling it to work effectively across diverse tasks and make optimal use of multiple references. In testing on the SummEval dataset, Fusion-Eval achieved a Spearman correlation of 0.96, outperforming other evaluators. The success of Fusion-Eval underscores the potential of LLMs to produce evaluations that closely align human perspectives, setting a new standard in the field of LLM evaluation.

Current role-play studies often rely on unvalidated LLM-as-a-judge paradigms, which may fail to reflect how humans perceive role fidelity. A key prerequisite for human-aligned evaluation is role identification, the ability to recognize who is speaking based on dialogue context. We argue that any meaningful judgment of role-playing quality (how well a character is played) fundamentally depends on first correctly attributing words and actions to the correct persona (who is speaking). We present PersonaEval, the first benchmark designed to test whether LLM evaluators can reliably identify human roles. PersonaEval uses human-authored dialogues from novels, scripts, and video transcripts, challenging models to determine the correct persona according to the conversation context. Our experiments, including a human study, show that even the best-performing LLMs reach only around 69% accuracy, well below the level needed for reliable evaluation. In contrast, human participants perform near ceiling with 90.8% accuracy, highlighting that current LLM evaluators are still not human enough to effectively judge role-play scenarios. To better understand this gap, we examine training-time adaptation and test-time compute, suggesting that reliable evaluation requires more than task-specific tuning, but depends on strong, human-like reasoning abilities in LLM evaluators. We release our benchmark at https://github.com/maple-zhou/PersonaEval.

As large-language models have been increasingly used as automatic raters for evaluating free-form content, including document summarization, dialog, and story generation, work has been dedicated to evaluating such models by measuring their correlations with human judgment. For sample-level performance, methods which operate by using pairwise comparisons between machine-generated text perform well but often lack the ability to assign absolute scores to individual summaries, an ability crucial for use cases that require thresholding. In this work, we propose a direct-scoring method which uses synthetic summaries to act as pairwise machine rankings at test time. We show that our method performs comparably to state-of-the-art pairwise evaluators in terms of axis-averaged sample-level correlations on the SummEval (+0.03), TopicalChat (-0.03), and HANNA (+0.05) meta-evaluation benchmarks, and release the synthetic in-context summaries as data to facilitate future work.

Recently, large language models (LLMs) have been deployed to tackle various software engineering (SE) tasks like code generation, significantly advancing the automation of SE tasks. However, assessing the quality of these LLM-generated code and text remains challenging. The commonly used Pass@k metric necessitates extensive unit tests and configured environments, demands a high labor cost, and is not suitable for evaluating LLM-generated text. Conventional metrics like BLEU, which measure only lexical rather than semantic similarity, have also come under scrutiny. In response, a new trend has emerged to employ LLMs for automated evaluation, known as LLM-as-a-judge. These LLM-as-a-judge methods are claimed to better mimic human assessment than conventional metrics without relying on high-quality reference answers. Nevertheless, their exact human alignment in SE tasks remains unexplored. In this paper, we empirically explore LLM-as-a-judge methods for evaluating SE tasks, focusing on their alignment with human judgments. We select seven LLM-as-a-judge methods that utilize general-purpose LLMs, alongside two LLMs specifically fine-tuned for evaluation. After generating and manually scoring LLM responses on three recent SE datasets of code translation, code generation, and code summarization, we then prompt these methods to evaluate each response. Finally, we compare the scores generated by these methods with human evaluation. The results indicate that output-based methods reach the highest Pearson correlation of 81.32 and 68.51 with human scores in code translation and generation, achieving near-human evaluation, noticeably outperforming ChrF++, one of the best conventional metrics, at 34.23 and 64.92. Such output-based methods prompt LLMs to output judgments directly, and exhibit more balanced score distributions that resemble human score patterns. Finally, we provide...

Scaling test-time computation, or affording a generator large language model (LLM) extra compute during inference, typically employs the help of external non-generative evaluators (i.e., reward models). Concurrently, LLM-judges, models trained to generate evaluations and critiques (explanations) in natural language, are becoming increasingly popular in automatic evaluation. Despite judge empirical successes, their effectiveness as evaluators in test-time scaling settings is largely unknown. In this paper, we introduce the Judge Evaluation for Test-Time Scaling (JETTS) benchmark, which evaluates judge performance in three domains (math reasoning, code generation, and instruction following) under three task settings: response reranking, step-level beam search, and critique-based response refinement. We evaluate 10 different judge models (7B-70B parameters) for 8 different base generator models (6.7B-72B parameters). Our benchmark shows that while judges are competitive with outcome reward models in reranking, they are consistently worse than process reward models in beam search procedures. Furthermore, though unique to LLM-judges, their natural language critiques are currently ineffective in guiding the generator towards better responses.

Multi-dimensional evaluation is the dominant paradigm for human evaluation in Natural Language Generation (NLG), i.e., evaluating the generated text from multiple explainable dimensions, such as coherence and fluency. However, automatic evaluation in NLG is still dominated by similarity-based metrics, and we lack a reliable framework for a more comprehensive evaluation of advanced models. In this paper, we propose a unified multi-dimensional evaluator UniEval for NLG. We re-frame NLG evaluation as a Boolean Question Answering (QA) task, and by guiding the model with different questions, we can use one evaluator to evaluate from multiple dimensions. Furthermore, thanks to the unified Boolean QA format, we are able to introduce an intermediate learning phase that enables UniEval to incorporate external knowledge from multiple related tasks and gain further improvement. Experiments on three typical NLG tasks show that UniEval correlates substantially better with human judgments than existing metrics. Specifically, compared to the top-performing unified evaluators, UniEval achieves a 23% higher correlation on text summarization, and over 43% on dialogue response generation. Also, UniEval demonstrates a strong zero-shot learning ability for unseen evaluation dimensions and tasks. Source code, data and all pre-trained evaluators are available on our GitHub repository (https://github.com/maszhongming/UniEval).

The performance of text summarization has been greatly boosted by pre-trained language models. A main concern of existing methods is that most generated summaries are not factually inconsistent with their source documents. To alleviate the problem, many efforts have focused on developing effective factuality evaluation metrics based on natural language inference, question answering, and syntactic dependency et al. However, these approaches are limited by either their high computational complexity or the uncertainty introduced by multi-component pipelines, resulting in only partial agreement with human judgement. Most recently, large language models(LLMs) have shown excellent performance in not only text generation but also language comprehension. In this paper, we particularly explore ChatGPT's ability to evaluate factual inconsistency under a zero-shot setting by examining it on both coarse-grained and fine-grained evaluation tasks including binary entailment inference, summary ranking, and consistency rating. Experimental results indicate that ChatGPT generally outperforms previous evaluation metrics across the three tasks, indicating its great potential for factual inconsistency evaluation. However, a closer inspection of ChatGPT's output reveals certain limitations including its preference for more lexically similar candidates, false reasoning, and inadequate understanding of instructions.

In this paper, we uncover a systematic bias in the evaluation paradigm of adopting large language models~(LLMs), e.g., GPT-4, as a referee to score and compare the quality of responses generated by candidate models. We find that the quality ranking of candidate responses can be easily hacked by simply altering their order of appearance in the context. This manipulation allows us to skew the evaluation result, making one model appear considerably superior to the other, e.g., Vicuna-13B could beat ChatGPT on 66 over 80 tested queries with ChatGPT as an evaluator. To address this issue, we propose a calibration framework with three simple yet effective strategies: 1) Multiple Evidence Calibration, which requires the evaluator model to generate multiple evaluation evidence before assigning ratings; 2) Balanced Position Calibration, which aggregates results across various orders to determine the final score; 3) Human-in-the-Loop Calibration, which introduces a balanced position diversity entropy to measure the difficulty of each example and seeks human assistance when needed. We also manually annotate the "win/tie/lose" outcomes of responses from ChatGPT and Vicuna-13B in the Vicuna Benchmark's question prompt, and extensive experiments demonstrate that our approach successfully mitigates evaluation bias, resulting in closer alignment with human judgments. We release our code and human annotation at https://github.com/i-Eval/FairEval to facilitate future research.

There is growing interest in systems that generate captions for scientific figures. However, assessing these systems output poses a significant challenge. Human evaluation requires academic expertise and is costly, while automatic evaluation depends on often low-quality author-written captions. This paper investigates using large language models (LLMs) as a cost-effective, reference-free method for evaluating figure captions. We first constructed SCICAP-EVAL, a human evaluation dataset that contains human judgments for 3,600 scientific figure captions, both original and machine-made, for 600 arXiv figures. We then prompted LLMs like GPT-4 and GPT-3 to score (1-6) each caption based on its potential to aid reader understanding, given relevant context such as figure-mentioning paragraphs. Results show that GPT-4, used as a zero-shot evaluator, outperformed all other models and even surpassed assessments made by Computer Science and Informatics undergraduates, achieving a Kendall correlation score of 0.401 with Ph.D. students rankings

Despite recent advances in text-to-3D generative methods, there is a notable absence of reliable evaluation metrics. Existing metrics usually focus on a single criterion each, such as how well the asset aligned with the input text. These metrics lack the flexibility to generalize to different evaluation criteria and might not align well with human preferences. Conducting user preference studies is an alternative that offers both adaptability and human-aligned results. User studies, however, can be very expensive to scale. This paper presents an automatic, versatile, and human-aligned evaluation metric for text-to-3D generative models. To this end, we first develop a prompt generator using GPT-4V to generate evaluating prompts, which serve as input to compare text-to-3D models. We further design a method instructing GPT-4V to compare two 3D assets according to user-defined criteria. Finally, we use these pairwise comparison results to assign these models Elo ratings. Experimental results suggest our metric strongly align with human preference across different evaluation criteria.

Recently, the emergence of ChatGPT has attracted wide attention from the computational linguistics community. Many prior studies have shown that ChatGPT achieves remarkable performance on various NLP tasks in terms of automatic evaluation metrics. However, the ability of ChatGPT to serve as an evaluation metric is still underexplored. Considering assessing the quality of natural language generation (NLG) models is an arduous task and NLG metrics notoriously show their poor correlation with human judgments, we wonder whether ChatGPT is a good NLG evaluation metric. In this report, we provide a preliminary meta-evaluation on ChatGPT to show its reliability as an NLG metric. In detail, we regard ChatGPT as a human evaluator and give task-specific (e.g., summarization) and aspect-specific (e.g., relevance) instruction to prompt ChatGPT to evaluate the generated results of NLG models. We conduct experiments on five NLG meta-evaluation datasets (including summarization, story generation and data-to-text tasks). Experimental results show that compared with previous automatic metrics, ChatGPT achieves state-of-the-art or competitive correlation with human judgments in most cases. In addition, we find that the effectiveness of the ChatGPT evaluator might be influenced by the creation method of the meta-evaluation datasets. For the meta-evaluation datasets which are created greatly depending on the reference and thus are biased, the ChatGPT evaluator might lose its effectiveness. We hope our preliminary study could prompt the emergence of a general-purposed reliable NLG metric.

The continuous advancement of large language models (LLMs) has brought increasing attention to the critical issue of developing fair and reliable methods for evaluating their performance. Particularly, the emergence of cheating phenomena, such as test set leakage and prompt format overfitting, poses significant challenges to the reliable evaluation of LLMs. As evaluation frameworks commonly use Regular Expression (RegEx) for answer extraction, models may adjust their responses to fit formats easily handled by RegEx. Nevertheless, the key answer extraction module based on RegEx frequently suffers from extraction errors. Furthermore, recent studies proposing fine-tuned LLMs as judge models for automated evaluation face challenges in terms of generalization ability and fairness. This paper comprehensively analyzes the entire LLM evaluation chain and demonstrates that optimizing the key answer extraction module improves extraction accuracy and enhances evaluation reliability. Our findings suggest that improving the key answer extraction module can lead to higher judgment accuracy and improved evaluation efficiency compared to the judge models. To address these issues, we propose xFinder, a novel evaluator for answer extraction and matching in LLM evaluation. As part of this process, we create a specialized dataset, the Key Answer Finder (KAF) dataset, to ensure effective model training and evaluation. Generalization tests and real-world evaluations show that the smallest xFinder model, with only 500 million parameters, achieves an average extraction accuracy of 93.42\%. In contrast, RegEx accuracy in the best evaluation framework is 74.38\%. The final judgment accuracy of xFinder reaches 97.61\%, outperforming existing evaluation frameworks and judge models.

Text-to-3D generation has achieved remarkable progress in recent years, yet evaluating these methods remains challenging for two reasons: i) Existing benchmarks lack fine-grained evaluation on different prompt categories and evaluation dimensions. ii) Previous evaluation metrics only focus on a single aspect (e.g., text-3D alignment) and fail to perform multi-dimensional quality assessment. To address these problems, we first propose a comprehensive benchmark named MATE-3D. The benchmark contains eight well-designed prompt categories that cover single and multiple object generation, resulting in 1,280 generated textured meshes. We have conducted a large-scale subjective experiment from four different evaluation dimensions and collected 107,520 annotations, followed by detailed analyses of the results. Based on MATE-3D, we propose a novel quality evaluator named HyperScore. Utilizing hypernetwork to generate specified mapping functions for each evaluation dimension, our metric can effectively perform multi-dimensional quality assessment. HyperScore presents superior performance over existing metrics on MATE-3D, making it a promising metric for assessing and improving text-to-3D generation. The project is available at https://mate-3d.github.io/.

Despite their promise to perform complex reasoning, large language models (LLMs) have been shown to have limited effectiveness in end-to-end planning. This has inspired an intriguing question: if these models cannot plan well, can they still contribute to the planning framework as a helpful plan evaluator? In this work, we generalize this question to consider LLMs augmented with visual understanding, i.e., Vision-Language Models (VLMs). We introduce PathEval, a novel benchmark evaluating VLMs as plan evaluators in complex path-planning scenarios. Succeeding in the benchmark requires a VLM to be able to abstract traits of optimal paths from the scenario description, demonstrate precise low-level perception on each path, and integrate this information to decide the better path. Our analysis of state-of-the-art VLMs reveals that these models face significant challenges on the benchmark. We observe that the VLMs can precisely abstract given scenarios to identify the desired traits and exhibit mixed performance in integrating the provided information. Yet, their vision component presents a critical bottleneck, with models struggling to perceive low-level details about a path. Our experimental results show that this issue cannot be trivially addressed via end-to-end fine-tuning; rather, task-specific discriminative adaptation of these vision encoders is needed for these VLMs to become effective path evaluators.

End-to-end spoken dialogue models such as GPT-4o-audio have recently garnered significant attention in the speech domain. However, the evaluation of spoken dialogue models' conversational performance has largely been overlooked. This is primarily due to the intelligent chatbots convey a wealth of non-textual information which cannot be easily measured using text-based language models like ChatGPT. To address this gap, we propose WavReward, a reward feedback model based on audio language models that can evaluate both the IQ and EQ of spoken dialogue systems with speech input. Specifically, 1) based on audio language models, WavReward incorporates the deep reasoning process and the nonlinear reward mechanism for post-training. By utilizing multi-sample feedback via the reinforcement learning algorithm, we construct a specialized evaluator tailored to spoken dialogue models. 2) We introduce ChatReward-30K, a preference dataset used to train WavReward. ChatReward-30K includes both comprehension and generation aspects of spoken dialogue models. These scenarios span various tasks, such as text-based chats, nine acoustic attributes of instruction chats, and implicit chats. WavReward outperforms previous state-of-the-art evaluation models across multiple spoken dialogue scenarios, achieving a substantial improvement about Qwen2.5-Omni in objective accuracy from 55.1% to 91.5%. In subjective A/B testing, WavReward also leads by a margin of 83%. Comprehensive ablation studies confirm the necessity of each component of WavReward. All data and code will be publicly at https://github.com/jishengpeng/WavReward after the paper is accepted.

Evaluating the conversational abilities of large language models (LLMs) remains a challenging task. Current mainstream approaches primarily rely on the ``LLM-as-a-judge" paradigm, where an LLM is prompted to serve as an evaluator to assess dialogue quality. However, such methods often suffer from various biases, which undermine the reliability and consistency of the evaluation results. To mitigate these biases, recent methods employ multiple LLMs as judges and aggregate their judgments to select the optimal assessment. Although effective, this multi-judge approach incurs significant computational overhead during inference. In this paper, we propose an efficient multi-turn dialogue evaluator that captures the collective wisdom of multiple LLM judges by aggregating their preference knowledge into a single model. Our approach preserves the advantages of diverse multi-judge feedback while drastically reducing the evaluation cost, enabling fast and flexible dialogue quality assessment. Extensive experiments on seven single rating and pairwise comparison dialogue evaluation benchmarks demonstrate that our method outperforms existing baselines across diverse scenarios, showcasing its efficiency and robustness.

Text evaluation has historically posed significant challenges, often demanding substantial labor and time cost. With the emergence of large language models (LLMs), researchers have explored LLMs' potential as alternatives for human evaluation. While these single-agent-based approaches show promise, experimental results suggest that further advancements are needed to bridge the gap between their current effectiveness and human-level evaluation quality. Recognizing that best practices of human evaluation processes often involve multiple human annotators collaborating in the evaluation, we resort to a multi-agent debate framework, moving beyond single-agent prompting strategies. The multi-agent-based approach enables a group of LLMs to synergize with an array of intelligent counterparts, harnessing their distinct capabilities and expertise to enhance efficiency and effectiveness in handling intricate tasks. In this paper, we construct a multi-agent referee team called ChatEval to autonomously discuss and evaluate the quality of generated responses from different models on open-ended questions and traditional natural language generation (NLG) tasks. Our analysis shows that ChatEval transcends mere textual scoring, offering a human-mimicking evaluation process for reliable assessments. Our code is available at https://github.com/chanchimin/ChatEval.

Large Language Models (LLMs) have recently been shown to be effective as automatic evaluators with simple prompting and in-context learning. In this work, we assemble 15 LLMs of four different size ranges and evaluate their output responses by preference ranking from the other LLMs as evaluators, such as System Star is better than System Square. We then evaluate the quality of ranking outputs introducing the Cognitive Bias Benchmark for LLMs as Evaluators (CoBBLEr), a benchmark to measure six different cognitive biases in LLM evaluation outputs, such as the Egocentric bias where a model prefers to rank its own outputs highly in evaluation. We find that LLMs are biased text quality evaluators, exhibiting strong indications on our bias benchmark (average of 40% of comparisons across all models) within each of their evaluations that question their robustness as evaluators. Furthermore, we examine the correlation between human and machine preferences and calculate the average Rank-Biased Overlap (RBO) score to be 49.6%, indicating that machine preferences are misaligned with humans. According to our findings, LLMs may still be unable to be utilized for automatic annotation aligned with human preferences. Our project page is at: https://minnesotanlp.github.io/cobbler.

Large language models (LLMs) have demonstrated impressive reasoning capabilities, yet there is ongoing debate about these abilities and the potential data contamination problem recently. This paper aims to evaluate the reasoning capacities of LLMs, specifically in solving recent competition-level programming problems in Codeforces, which are expert-crafted and unique, requiring deep understanding and robust reasoning skills. We first provide a comprehensive evaluation of GPT-4's peiceived zero-shot performance on this task, considering various aspects such as problems' release time, difficulties, and types of errors encountered. Surprisingly, the peiceived performance of GPT-4 has experienced a cliff like decline in problems after September 2021 consistently across all the difficulties and types of problems, which shows the potential data contamination, as well as the challenges for any existing LLM to solve unseen complex reasoning problems. We further explore various approaches such as fine-tuning, Chain-of-Thought prompting and problem description simplification, unfortunately none of them is able to consistently mitigate the challenges. Through our work, we emphasis the importance of this excellent data source for assessing the genuine reasoning capabilities of LLMs, and foster the development of LLMs with stronger reasoning abilities and better generalization in the future.

Employing Large Language Models (LLMs) to assess the quality of generated responses, such as prompting instruct-tuned models or fine-tuning judge models, has become a widely adopted evaluation method. It is also known that such evaluators are vulnerable to biases, such as favoring longer responses. While it is important to overcome this problem, the specifics of these biases remain under-explored. In this work, we qualitatively identify six types of biases inherent in various judge models. We propose EvalBiasBench as a meta-evaluation collection of hand-crafted test cases for each bias type. Additionally, we present de-biasing dataset construction methods and the associated preference dataset OffsetBias. Experimental results demonstrate that fine-tuning on our dataset significantly enhances the robustness of judge models against biases and improves performance across most evaluation scenarios. We release our datasets and the fine-tuned judge model to public.

Retrieval-Augmented Generation (RAG) has emerged as a common paradigm to use Large Language Models (LLMs) alongside private and up-to-date knowledge bases. In this work, we address the challenges of using LLM-as-a-Judge when evaluating grounded answers generated by RAG systems. To assess the calibration and discrimination capabilities of judge models, we identify 7 generator failure modes and introduce GroUSE (Grounded QA Unitary Scoring of Evaluators), a meta-evaluation benchmark of 144 unit tests. This benchmark reveals that existing automated RAG evaluation frameworks often overlook important failure modes, even when using GPT-4 as a judge. To improve on the current design of automated RAG evaluation frameworks, we propose a novel pipeline and find that while closed models perform well on GroUSE, state-of-the-art open-source judges do not generalize to our proposed criteria, despite strong correlation with GPT-4's judgement. Our findings suggest that correlation with GPT-4 is an incomplete proxy for the practical performance of judge models and should be supplemented with evaluations on unit tests for precise failure mode detection. We further show that finetuning Llama-3 on GPT-4's reasoning traces significantly boosts its evaluation capabilities, improving upon both correlation with GPT-4's evaluations and calibration on reference situations.

Large Language Models (LLMs) are increasingly serving as evaluators in Natural Language Generation (NLG) tasks. However, the capabilities of LLMs in scoring NLG quality remain inadequately explored. Current studies depend on human assessments and simple metrics that fail to capture the discernment of LLMs across diverse NLG tasks. To address this gap, we propose the Discernment of Hierarchical Perturbation (DHP) benchmarking framework, which provides quantitative discernment scores for LLMs utilizing hierarchically perturbed text data and statistical tests to measure the NLG evaluation capabilities of LLMs systematically. We have re-established six evaluation datasets for this benchmark, covering four NLG tasks: Summarization, Story Completion, Question Answering, and Translation. Our comprehensive benchmarking of five major LLM series provides critical insight into their strengths and limitations as NLG evaluators.

Recent advancements in large reasoning models (LRMs) have introduced an intermediate "thinking" process prior to generating final answers, improving their reasoning capabilities on complex downstream tasks. However, the potential of LRMs as evaluators for machine translation (MT) quality remains underexplored. We provides the first systematic analysis of LRM-as-a-judge in MT evaluation. We identify key challenges, revealing LRMs require tailored evaluation materials, tend to "overthink" simpler instances and have issues with scoring mechanisms leading to overestimation. To address these, we propose to calibrate LRM thinking by training them on synthetic, human-like thinking trajectories. Our experiments on WMT24 Metrics benchmarks demonstrate that this approach largely reduces thinking budgets by ~35x while concurrently improving evaluation performance across different LRM scales from 7B to 32B (e.g., R1-Distill-Qwen-7B achieves a +8.7 correlation point improvement). These findings highlight the potential of efficiently calibrated LRMs to advance fine-grained automatic MT evaluation.

Since the adoption of large language models (LLMs) for text evaluation has become increasingly prevalent in the field of natural language processing (NLP), a series of existing works attempt to optimize the prompts for LLM evaluators to improve their alignment with human judgment. However, their efforts are limited to optimizing individual factors of evaluation prompts, such as evaluation criteria or output formats, neglecting the combinatorial impact of multiple factors, which leads to insufficient optimization of the evaluation pipeline. Nevertheless, identifying well-behaved prompting strategies for adjusting multiple factors requires extensive enumeration. To this end, we comprehensively integrate 8 key factors for evaluation prompts and propose a novel automatic prompting strategy optimization method called Heuristic Prompting Strategy Search (HPSS). Inspired by the genetic algorithm, HPSS conducts an iterative search to find well-behaved prompting strategies for LLM evaluators. A heuristic function is employed to guide the search process, enhancing the performance of our algorithm. Extensive experiments across four evaluation tasks demonstrate the effectiveness of HPSS, consistently outperforming both human-designed evaluation prompts and existing automatic prompt optimization methods. Our code is available at https://github.com/thu-coai/HPSS.

Large language models (LLMs) have emerged as a promising alternative to expensive human evaluations. However, the alignment and coverage of LLM-based evaluations are often limited by the scope and potential bias of the evaluation prompts and criteria. To address this challenge, we propose HD-Eval, a novel framework that iteratively aligns LLM-based evaluators with human preference via Hierarchical Criteria Decomposition. HD-Eval inherits the essence from the evaluation mindset of human experts and enhances the alignment of LLM-based evaluators by decomposing a given evaluation task into finer-grained criteria, aggregating them according to estimated human preferences, pruning insignificant criteria with attribution, and further decomposing significant criteria. By integrating these steps within an iterative alignment training process, we obtain a hierarchical decomposition of criteria that comprehensively captures aspects of natural language at multiple levels of granularity. Implemented as a white box, the human preference-guided aggregator is efficient to train and more explainable than relying solely on prompting, and its independence from model parameters makes it applicable to closed-source LLMs. Extensive experiments on three evaluation domains demonstrate the superiority of HD-Eval in further aligning state-of-the-art evaluators and providing deeper insights into the explanation of evaluation results and the task itself.

Automated Text-to-Speech (TTS) evaluation methods (Mean Opinion Score (MOS) predictors and Audio Large Language Models (Audio-LLM) judges) are expected to reflect human perception, yet it is unclear how well they capture the distinct aspects of speech that listeners actually perceive. We deconstruct "naturalness" into a linguistically grounded annotation schema spanning 10 distinct perceptual dimensions, and use it to construct the first dimension-level meta-evaluation benchmark for TTS, comprising 860 utterances annotated by trained linguist raters. Results from benchmarking four MOS predictors and four Audio-LLM judges reveal that MOS predictors collapse onto acoustic signal quality, while Audio-LLM judges show selective, prompt-dependent detection that does not generalise across all dimensions. Neither class reliably captures a breadth of linguistically structured speech errors. Our dataset, annotation schema, and evaluation code are publicly released to support more targeted and interpretable TTS evaluation.

Evaluation of multilingual Large Language Models (LLMs) is challenging due to a variety of factors -- the lack of benchmarks with sufficient linguistic diversity, contamination of popular benchmarks into LLM pre-training data and the lack of local, cultural nuances in translated benchmarks. In this work, we study human and LLM-based evaluation in a multilingual, multi-cultural setting. We evaluate 30 models across 10 Indic languages by conducting 90K human evaluations and 30K LLM-based evaluations and find that models such as GPT-4o and Llama-3 70B consistently perform best for most Indic languages. We build leaderboards for two evaluation settings - pairwise comparison and direct assessment and analyse the agreement between humans and LLMs. We find that humans and LLMs agree fairly well in the pairwise setting but the agreement drops for direct assessment evaluation especially for languages such as Bengali and Odia. We also check for various biases in human and LLM-based evaluation and find evidence of self-bias in the GPT-based evaluator. Our work presents a significant step towards scaling up multilingual evaluation of LLMs.

The impressive performance of large language models (LLMs) has attracted considerable attention from the academic and industrial communities. Besides how to construct and train LLMs, how to effectively evaluate and compare the capacity of LLMs has also been well recognized as an important yet difficult problem. Existing paradigms rely on either human annotators or model-based evaluators to evaluate the performance of LLMs on different tasks. However, these paradigms often suffer from high cost, low generalizability, and inherited biases in practice, which make them incapable of supporting the sustainable development of LLMs in long term. In order to address these issues, inspired by the peer review systems widely used in academic publication process, we propose a novel framework that can automatically evaluate LLMs through a peer-review process. Specifically, for the evaluation of a specific task, we first construct a small qualification exam to select "reviewers" from a couple of powerful LLMs. Then, to actually evaluate the "submissions" written by different candidate LLMs, i.e., the evaluatees, we use the reviewer LLMs to rate or compare the submissions. The final ranking of evaluatee LLMs is generated based on the results provided by all reviewers. We conducted extensive experiments on text summarization tasks with eleven LLMs including GPT-4. The results demonstrate the existence of biasness when evaluating using a single LLM. Also, our PRE model outperforms all the baselines, illustrating the effectiveness of the peer review mechanism.

As language model (LM) outputs get more and more natural, it is becoming more difficult than ever to evaluate their quality. Simultaneously, increasing LMs' "thinking" time through scaling test-time compute has proven an effective technique to solve challenging problems in domains such as math and code. This raises a natural question: can an LM's evaluation capability also be improved by spending more test-time compute? To answer this, we investigate employing reasoning models-LMs that natively generate long chain-of-thought reasoning-as evaluators. Specifically, we examine methods to leverage more test-time compute by (1) using reasoning models, and (2) prompting these models to evaluate not only the response as a whole (i.e., outcome evaluation) but also assess each step in the response separately (i.e., process evaluation). In experiments, we observe that the evaluator's performance improves monotonically when generating more reasoning tokens, similar to the trends observed in LM-based generation. Furthermore, we use these more accurate evaluators to rerank multiple generations, and demonstrate that spending more compute at evaluation time can be as effective as using more compute at generation time in improving an LM's problem-solving capability.

Agent-based editing models have substantially advanced interactive experiences, processing quality, and creative flexibility. However, two critical challenges persist: (1) instruction hallucination, text-only chain-of-thought (CoT) reasoning cannot fully prevent factual errors due to inherent information bottlenecks; (2) reward hacking, dynamic policy optimization against static reward models allows agents to exploit flaws in reward functions. To address these issues, we propose JarvisEvo, a unified image editing agent that emulates an expert human designer by iteratively editing, selecting appropriate tools, evaluating results, and reflecting on its own decisions to refine outcomes. JarvisEvo offers three key advantages: (1) an interleaved multimodal chain-of-thought (iMCoT) reasoning mechanism that enhances instruction following and editing quality; (2) a synergistic editor-evaluator policy optimization (SEPO) framework that enables self-improvement without external rewards, effectively mitigating reward hacking; and (3) support for both global and local fine-grained editing through seamless integration of Adobe Lightroom. On ArtEdit-Bench, JarvisEvo outperforms Nano-Banana by an average of 18.95% on preservative editing metrics, including a substantial 44.96% improvement in pixel-level content fidelity. Project page: https://jarvisevo.vercel.app/

As audio-visual generative models evolve into world simulators, cross-modal synchronization stands as a critical proxy for assessing the consistency of world dynamics and causality in generated content. However, existing evaluation metrics presume structural correctness, reducing synchronization to mere temporal alignment. Consequently, they fail on generative outputs, especially when exhibiting structural hallucinations and asymmetric cross-modal relations, which currently mandate expert human annotation to assess synchronization. This dependency introduces a critical paradox: human evaluators rely on relative, reference-dependent comparisons, whereas automated metrics require reference-free, absolute scalars. We resolve this paradox by proposing a framework that distills relative human perception into a continuous, globally consistent metric. First, we introduce SynthSync, a dataset of generative failures ranked via pairwise human annotations. Second, we adapt the Omni-LLM equipped with a continuous latent projection to translate relative human rankings into continuous absolute values. Third, we propose Real-Valued Group Relative Policy Optimization (R-GRPO) to internalize the global causal structure of synchronization via listwise score distributions. Empirically, our metric achieves state-of-the-art human preference alignment. We leverage this estimator to establish a standardized benchmark, advancing AV-Gen assessment from low-level signal correlation to visually grounded causality.

Recent advancements in the field of natural language generation have facilitated the use of large language models to assess the quality of generated text. Although these models have shown promising results in tasks such as machine translation and summarization, their applicability in code generation tasks remains limited without human involvement. The complexity of programming concepts required for such tasks makes it difficult to develop evaluation metrics that align with human judgment. Token-matching-based metrics, such as BLEU, have demonstrated weak correlations with human practitioners in code generation tasks. Moreover, the utilization of human-written test suites to evaluate functional correctness can be challenging in domains with low resources. To overcome these obstacles, we propose a new evaluation framework based on the GPT-3.5 (GPT-3.5-turbo), for code generation assessments. Our framework addresses the limitations of existing approaches by achieving superior correlations with functional correctness and human preferences, without the need for test oracles or references. We evaluate the efficacy of our framework on two different tasks and four programming languages, comparing its performance with the state-of-the-art CodeBERTScore metric, which relies on a pre-trained model. Our results demonstrate that our framework surpasses CodeBERTScore, delivering high levels of accuracy and consistency across various programming languages and tasks. We also make our evaluation framework and datasets available to the public at https://github.com/terryyz/llm-code-eval, encouraging further research in the evaluation of code generation.

Large language models (LLMs) have shown remarkable ability on controllable text generation. However, the potential of LLMs in generating text from structured tables remains largely under-explored. In this paper, we study the capabilities of LLMs for table-to-text generation tasks, particularly aiming to investigate their performance in generating natural language statements that can be logically entailed by a provided table. First, we investigate how LLMs compare to state-of-the-art table-to-text fine-tuned models, and demonstrate that LLMs can generate statements with higher faithfulness compared with previous state-of-the-art fine-tuned models. Given this finding, we next explore whether LLMs can serve as faithfulness-level automated evaluation metrics. Through human evaluation, we show that evaluation metrics adopted from LLMs correlates better with human judgments compared with existing faithfulness-level metrics. Finally, we demonstrate that LLMs using chain-of-thought prompting can generate high-fidelity natural language feedback for other table-to-text models' generations, provide insights for future work regarding the distillation of text generation capabilities from LLMs to smaller models.

Recent studies show that sentence-level extractive QA, i.e., based on Answer Sentence Selection (AS2), is outperformed by Generation-based QA (GenQA) models, which generate answers using the top-k answer sentences ranked by AS2 models (a la retrieval-augmented generation style). In this paper, we propose a novel training paradigm for GenQA using supervision from automatic QA evaluation models (GAVA). Specifically, we propose three strategies to transfer knowledge from these QA evaluation models to a GenQA model: (i) augmenting training data with answers generated by the GenQA model and labelled by GAVA (either statically, before training, or (ii) dynamically, at every training epoch); and (iii) using the GAVA score for weighting the generator loss during the learning of the GenQA model. We evaluate our proposed methods on two academic and one industrial dataset, obtaining a significant improvement in answering accuracy over the previous state of the art.

Large language models (LLMs) have shown promise as automated evaluators for assessing the quality of answers generated by AI systems. However, these LLM-based evaluators exhibit position bias, or inconsistency, when used to evaluate candidate answers in pairwise comparisons, favoring either the first or second answer regardless of content. To address this limitation, we propose PORTIA, an alignment-based system designed to mimic human comparison strategies to calibrate position bias in a lightweight yet effective manner. Specifically, PORTIA splits the answers into multiple segments, aligns similar content across candidate answers, and then merges them back into a single prompt for evaluation by LLMs. We conducted extensive experiments with six diverse LLMs to evaluate 11,520 answer pairs. Our results show that PORTIA markedly enhances the consistency rates for all the models and comparison forms tested, achieving an average relative improvement of 47.46%. Remarkably, PORTIA enables less advanced GPT models to achieve 88% agreement with the state-of-the-art GPT-4 model at just 10% of the cost. Furthermore, it rectifies around 80% of the position bias instances within the GPT-4 model, elevating its consistency rate up to 98%. Subsequent human evaluations indicate that the PORTIA-enhanced GPT-3.5 model can even surpass the standalone GPT-4 in terms of alignment with human evaluators. These findings highlight PORTIA's ability to correct position bias, improve LLM consistency, and boost performance while keeping cost-efficiency. This represents a valuable step toward a more reliable and scalable use of LLMs for automated evaluations across diverse applications.

We describe GEMBA, a GPT-based metric for assessment of translation quality, which works both with a reference translation and without. In our evaluation, we focus on zero-shot prompting, comparing four prompt variants in two modes, based on the availability of the reference. We investigate nine versions of GPT models, including ChatGPT and GPT-4. We show that our method for translation quality assessment only works with GPT~3.5 and larger models. Comparing to results from WMT22's Metrics shared task, our method achieves state-of-the-art accuracy in both modes when compared to MQM-based human labels. Our results are valid on the system level for all three WMT22 Metrics shared task language pairs, namely English into German, English into Russian, and Chinese into English. This provides a first glimpse into the usefulness of pre-trained, generative large language models for quality assessment of translations. We publicly release all our code and prompt templates used for the experiments described in this work, as well as all corresponding scoring results, to allow for external validation and reproducibility.

Past analyses of reinforcement learning from human feedback (RLHF) assume that the human fully observes the environment. What happens when human feedback is based only on partial observations? We formally define two failure cases: deception and overjustification. Modeling the human as Boltzmann-rational w.r.t. a belief over trajectories, we prove conditions under which RLHF is guaranteed to result in policies that deceptively inflate their performance, overjustify their behavior to make an impression, or both. To help address these issues, we mathematically characterize how partial observability of the environment translates into (lack of) ambiguity in the learned return function. In some cases, accounting for partial observability makes it theoretically possible to recover the return function and thus the optimal policy, while in other cases, there is irreducible ambiguity. We caution against blindly applying RLHF in partially observable settings and propose research directions to help tackle these challenges.

Large Language Models (LLMs) have demonstrated promising capabilities as automatic evaluators in assessing the quality of generated natural language. However, LLMs still exhibit biases in evaluation and often struggle to generate coherent evaluations that align with human assessments. In this work, we first conduct a systematic study of the misalignment between LLM evaluators and human judgement, revealing that existing calibration methods aimed at mitigating biases are insufficient for effectively aligning LLM evaluators. Inspired by the use of preference data in RLHF, we formulate the evaluation as a ranking problem and introduce Pairwise-preference Search (PairS), an uncertainty-guided search method that employs LLMs to conduct pairwise comparisons and efficiently ranks candidate texts. PairS achieves state-of-the-art performance on representative evaluation tasks and demonstrates significant improvements over direct scoring. Furthermore, we provide insights into the role of pairwise preference in quantifying the transitivity of LLMs and demonstrate how PairS benefits from calibration.

LLM-based auto-annotators have become a key component of the LLM development process due to their cost-effectiveness and scalability compared to human-based evaluation. However, these auto-annotators can introduce complex biases that are hard to remove. Even simple, known confounders such as preference for longer outputs remain in existing automated evaluation metrics. We propose a simple regression analysis approach for controlling biases in auto-evaluations. As a real case study, we focus on reducing the length bias of AlpacaEval, a fast and affordable benchmark for chat LLMs that uses LLMs to estimate response quality. Despite being highly correlated with human preferences, AlpacaEval is known to favor models that generate longer outputs. We introduce a length-controlled AlpacaEval that aims to answer the counterfactual question: "What would the preference be if the model's and baseline's output had the same length?". To achieve this, we first fit a generalized linear model to predict the biased output of interest (auto-annotator preferences) based on the mediators we want to control for (length difference) and other relevant features. We then obtain length-controlled preferences by predicting preferences while conditioning the GLM with a zero difference in lengths. Length-controlling not only improves the robustness of the metric to manipulations in model verbosity, we also find that it increases the Spearman correlation with LMSYS' Chatbot Arena from 0.94 to 0.98. We release the code and leaderboard at https://tatsu-lab.github.io/alpaca_eval/ .

Recent advances in large language models (LLMs) show the potential of using LLMs as evaluators for assessing the quality of text generations from LLMs. However, applying LLM evaluators naively to compare or judge between different systems can lead to unreliable results due to the intrinsic win rate estimation bias of LLM evaluators. In order to mitigate this problem, we propose two calibration methods, Bayesian Win Rate Sampling (BWRS) and Bayesian Dawid-Skene, both of which leverage Bayesian inference to more accurately infer the true win rate of generative language models. We empirically validate our methods on six datasets covering story generation, summarization, and instruction following tasks. We show that both our methods are effective in improving the accuracy of win rate estimation using LLMs as evaluators, offering a promising direction for reliable automatic text quality evaluation.

Measuring the quality of responses generated by LLMs is a challenging task, particularly when it comes to evaluating whether the response is aligned with human preference. A novel approach involves using the LLM itself to make evaluation and stabilizing the results through multiple independent evaluations, similar to a single-layer narrow LLM network. This network consists of a fixed number of neurons, with each neuron being the same LLM. In this paper, we draw upon the extensive research on deep neural networks to explore whether deeper and wider networks can lead to fairer evaluations. Specifically, inspired by the observation that different neurons in a neural network are responsible for detecting different concepts, we first adaptively generate as many neuron roles as possible for each evaluation sample. Each perspective corresponds to the role of a specific LLM neuron in the first layer. In subsequent layers, we follow the idea that higher layers in deep networks are responsible for more comprehensive features, each layer receives representations from all neurons in the previous layer, integrating the locally learned evaluation information to obtain a more comprehensive evaluation result. Interestingly, this network design resembles the process of academic paper reviewing. To validate the effectiveness of our method, we construct the largest and most diverse English evaluation benchmark LLMEval^2 for LLM evaluators, comprising 15 tasks, 8 abilities, and 2,553 samples. Experimental results demonstrate that a wider network (involving many reviewers) with 2 layers (one round of discussion) performs the best, improving kappa correlation coefficient from 0.28 to 0.34. We also leverage WideDeep to aid in the assessment of Chinese LLMs, which has accelerated the evaluation time by 4.6 times, resulting in a 60% cost saving. WideDeep achieves a remarkable 93% agreement level among humans.

Automatic evaluation with large language models, commonly known as LLM-as-a-judge, is now standard across reasoning and alignment tasks. Despite evaluating many samples in deployment, these evaluators typically (i) treat each case independently, missing the opportunity to accumulate experience, and (ii) rely on a single fixed prompt for all cases, neglecting the need for sample-specific evaluation criteria. We introduce Learning While Evaluating (LWE), a framework that allows evaluators to improve sequentially at inference time without requiring training or validation sets. LWE maintains an evolving meta-prompt that (i) produces sample-specific evaluation instructions and (ii) refines itself through self-generated feedback. Furthermore, we propose Selective LWE, which updates the meta-prompt only on self-inconsistent cases, focusing computation where it matters most. This selective approach retains the benefits of sequential learning while being far more cost-effective. Across two pairwise comparison benchmarks, Selective LWE outperforms strong baselines, empirically demonstrating that evaluators can improve during sequential testing with a simple selective update, learning most from the cases they struggle with.

We present GSM-MC, a multiple-choice (MC) dataset constructed by collecting answers and incorrect predictions on GSM8K from 60 open-source models. Through extensive experiments, we show that LLMs' performance on the MC version of this popular benchmark is strongly correlated with their performance on the original version and is quite robust to distractor choices and option orders, while the evaluation time is reduced by a factor of up to 30. Following similar procedures, we introduce MATH-MC, constructed from MATH, and PythonIO, a new program reasoning MC dataset constructed from HumanEval and MBPP. Experimental results indicate that LLMs' performance on these MC benchmarks leaves much room for improvement. Our data and code are available at https://github.com/Geralt-Targaryen/MC-Evaluation.

Self-improving agents are state-of-the-art (SOTA) on agentic coding benchmarks and have recently been extended to general domains. However, their search methods generally assume a stationary evaluation criterion: a fixed verifier, benchmark, or labeled dataset that remains valid as the agent improves. This ignores a central feature of evolution: species adapt as their environments change with them. We aim to bring the same principle to recursive self-improvement, making evaluation part of the improvement loop and opening search to evolving evaluators, adversarial objectives, and dynamic utilities that may surpass static benchmarks. We introduce the Red Queen Godel Machine (RQGM), an evolutionary framework for recursive self-improvement under non-stationary utilities. The RQGM makes this possible through controlled utility evolution: search is organized into epochs with a fixed within-epoch evaluation criterion, while the utility can be updated at epoch boundaries, so self-improvement guarantees hold per epoch as the objective evolves across them. We begin by showing that even on verifiable coding tasks, the RQGM improves test pass rate over the prior SOTA by adding a complementary agent-as-a-judge code-review signal. This signal is cheaper and the RQGM uses 1.35x-1.72x fewer tokens. We then turn to scientific paper writing and reviewing, and Olympiad-level proof writing and grading, where the RQGM improves performance over prior self-improving agents: co-evolved writers reach 1.78x-1.86x higher acceptance rates under a diverse agent-as-a-judge panel, while co-evolved graders reach 9% higher ground-truth accuracy. In paper reviewing, the strongest baseline reviewer over-accepts AI-generated papers at up to 1.91x the human rate. The RQGM corrects this by introducing an adversarial objective that discovers reviewers equally stringent on AI and human work.

Creativity evaluation remains a challenging frontier for large language models (LLMs). Current evaluations heavily rely on inefficient and costly human judgments, hindering progress in enhancing machine creativity. While automated methods exist, ranging from psychological testing to heuristic- or prompting-based approaches, they often lack generalizability or alignment with human judgment. To address these issues, in this paper, we propose a novel pairwise-comparison framework for assessing textual creativity, leveraging shared contextual instructions to improve evaluation consistency. We introduce CreataSet, a large-scale dataset with 100K+ human-level and 1M+ synthetic creative instruction-response pairs spanning diverse open-domain tasks. Through training on CreataSet, we develop an LLM-based evaluator named CrEval. CrEval demonstrates remarkable superiority over existing methods in alignment with human judgments. Experimental results underscore the indispensable significance of integrating both human-generated and synthetic data in training highly robust evaluators, and showcase the practical utility of CrEval in boosting the creativity of LLMs. We will release all data, code, and models publicly soon to support further research.

Automatic evaluation is an integral aspect of dialogue system research. The traditional reference-based NLG metrics are generally found to be unsuitable for dialogue assessment. Consequently, recent studies have suggested various unique, reference-free neural metrics that better align with human evaluations. Notably among them, large language models (LLMs), particularly the instruction-tuned variants like ChatGPT, are shown to be promising substitutes for human judges. Yet, existing works on utilizing LLMs for automatic dialogue evaluation are limited in their scope in terms of the number of meta-evaluation datasets, mode of evaluation, coverage of LLMs, etc. Hence, it remains inconclusive how effective these LLMs are. To this end, we conduct a comprehensive study on the application of LLMs for automatic dialogue evaluation. Specifically, we analyze the multi-dimensional evaluation capability of 30 recently emerged LLMs at both turn and dialogue levels, using a comprehensive set of 12 meta-evaluation datasets. Additionally, we probe the robustness of the LLMs in handling various adversarial perturbations at both turn and dialogue levels. Finally, we explore how model-level and dimension-level ensembles impact the evaluation performance. All resources are available at https://github.com/e0397123/comp-analysis.

As the capabilities of chatbots and their underlying LLMs continue to dramatically improve, evaluating their performance has increasingly become a major blocker to their further development. A major challenge is the available benchmarking datasets, which are largely static, outdated, and lacking in multilingual coverage, limiting their ability to capture subtle linguistic and cultural variations. This paper introduces MEDAL, an automated multi-agent framework for generating, evaluating, and curating more representative and diverse open-domain dialogue evaluation benchmarks. Our approach leverages several state-of-the-art LLMs to generate user-chatbot multilingual dialogues, conditioned on varied seed contexts. A strong LLM (GPT-4.1) is then used for a multidimensional analysis of the performance of the chatbots, uncovering noticeable cross-lingual performance differences. Guided by this large-scale evaluation, we curate a new meta-evaluation multilingual benchmark and human-annotate samples with nuanced quality judgments. This benchmark is then used to assess the ability of several reasoning and non-reasoning LLMs to act as evaluators of open-domain dialogues. We find that current LLMs struggle to detect nuanced issues, particularly those involving empathy and reasoning.

Despite the utility of Large Language Models (LLMs) across a wide range of tasks and scenarios, developing a method for reliably evaluating LLMs across varied contexts continues to be challenging. Modern evaluation approaches often use LLMs to assess responses generated by LLMs. However, the meta-evaluation conducted to assess the effectiveness of these LLMs as evaluators is typically constrained by the coverage of existing benchmarks or requires extensive human annotation. This underscores the urgency of methods for scalable meta-evaluation that can effectively, reliably, and efficiently evaluate the performance of LLMs as evaluators across diverse tasks and scenarios, particularly in potentially new, user-defined scenarios. To fill this gap, we propose ScaleEval, an agent-debate-assisted meta-evaluation framework that leverages the capabilities of multiple communicative LLM agents. This framework supports multi-round discussions to assist human annotators in discerning the most capable LLMs as evaluators, which significantly eases their workload in cases that used to require large-scale annotations during meta-evaluation. We release the code for our framework, which is publicly available at: https://github.com/GAIR-NLP/scaleeval.

Online misinformation poses a global risk with significant real-world consequences. To combat misinformation, current research relies on professionals like journalists and fact-checkers for annotating and debunking misinformation, and develops automated machine learning methods for detecting misinformation. Complementary to these approaches, recent research has increasingly concentrated on utilizing the power of ordinary social media users, a.k.a. "crowd", who act as eyes-on-the-ground proactively questioning and countering misinformation. Notably, recent studies show that 96% of counter-misinformation responses originate from them. Acknowledging their prominent role, we present the first systematic and comprehensive survey of research papers that actively leverage the crowds to combat misinformation. We first identify 88 papers related to crowd-based efforts, following a meticulous annotation process adhering to the PRISMA framework. We then present key statistics related to misinformation, counter-misinformation, and crowd input in different formats and topics. Upon holistic analysis of the papers, we introduce a novel taxonomy of the roles played by the crowds: (i)annotators who actively identify misinformation; (ii)evaluators who assess counter-misinformation effectiveness; (iii)creators who create counter-misinformation. This taxonomy explores the crowd's capabilities in misinformation detection, identifies prerequisites for effective counter-misinformation, and analyzes crowd-generated counter-misinformation. Then, we delve into (i)distinguishing individual, collaborative, and machine-assisted labeling for annotators; (ii)analyzing the effectiveness of counter-misinformation through surveys, interviews, and in-lab experiments for evaluators; and (iii)characterizing creation patterns and creator profiles for creators. Finally, we outline potential future research in this field.

LLMs-as-judges are the only way to evaluate safety at scale. Despite their importance, LLM-judges themselves are rarely evaluated beyond human agreement in simple, static benchmarks. We therefore investigate two under-explored but crucial properties of LLMs-as-judges: their susceptibility to relying on in context-information, and their steerability to differing safety definitions, which may not align with their internal safety priors. We evaluate the safety judging abilities of many generalist LLMs and safety-specific judges, and investigate the impact of task demonstrations, novel in-context information, and changing safety definitions. We find that while LLM-judges can learn from new information, they are broadly unlikely to adjust their evaluations if the context or safety definition contradicts their prior.

Video-language models and video agents can produce hallucinations that conflict with spatiotemporal evidence. Existing benchmarks mainly evaluate model hallucinations, and heterogeneous mechanisms make detector reliability difficult to compare. We introduce VidHalLoc, a benchmark that evaluates hallucination detection methods under a unified diagnostic evaluation protocol using 2,000 adversarial hallucination samples across Video Question Answering and Video Captioning tasks, spanning Ontology and Dynamic hallucination categories. To construct VidHalLoc efficiently, we introduce VideoHALO, a Harness Engineering-informed multi-agent workflow that decomposes data construction into four executable stages supported by a memory system and a communication protocol. Evaluation of fifteen methods reveals that the four dedicated detectors peak at an Overall accuracy of only 34.63%, indicating limited reliability across video hallucination types [Dataset Repository: https://huggingface.co/datasets/wesfggfd/VidHalLoc].

Most existing chain-of-thought (CoT) prompting methods suffer from the issues of generalizability and consistency, as they often rely on instance-specific solutions that may not be applicable to other cases and lack task-level consistency in their reasoning steps. To address these limitations, we propose a comprehensive framework, StrategyLLM, harnessing the capabilities of LLMs to construct generalizable and consistent few-shot prompts for various tasks automatically. To this end, StrategyLLM employs four LLM-based agents: strategy generator, executor, optimizer, and evaluator, working together to generate, evaluate, and select promising strategies for a given task. The experimental results demonstrate that StrategyLLM outperforms the competitive baseline CoT-SC that requires human-annotated solutions on 13 datasets across 4 challenging tasks without human involvement, including math reasoning (34.21% rightarrow 38.79%), commonsense reasoning (70.3% rightarrow 72.5%), algorithmic reasoning (51.7% rightarrow 62.0%), and symbolic reasoning (30.0% rightarrow 79.2%).

Intraoperative anesthesia requires systems to interpret evolving multimodal evidence, recommend timely management, and revise decisions as patient states change, yet existing benchmarks usually isolate perception or single-point reasoning. We introduce AnesTRACE, an evaluation suite comprising AnesTRACE-Bench and AnesTRACE-Eval. Built from public perioperative datasets with anesthesiologist annotation, AnesTRACE-Bench evaluates Intraoperative Perception, Single-point Anesthesia Decision-Making, and Multi-step Anesthesia Decision-Making. AnesTRACE-Eval assesses open-ended responses through anesthesiologist-defined criteria for Clinical Correctness, Evidence Grounding, Task Completeness, and Safety, with Temporal Consistency for multi-step decisions; its domain-specific evaluator is trained by supervised fine-tuning and preference alignment on expert-reviewed judgments. Across more than 30 models, fine-grained visual grounding and intervention selection remain difficult: the leading model reaches only 32.2 mIoU for TEE visual grounding and retains a 17.5\% Major/Critical Safety Error Rate in multi-step management. Evaluator alignment with anesthesiologists improves across both training stages, while the best decision quality is accompanied by a 74.3-second P95 Latency. These results show that aggregate performance alone does not establish safe, timely longitudinal decision-making. We release our code at https://zjudbxai.github.io/AnesTRACE/.

Current medical question answering systems have difficulty processing long, detailed and informally worded questions submitted by patients, called Consumer Health Questions (CHQs). To address this issue, we introduce a medical question understanding and answering system with knowledge grounding and semantic self-supervision. Our system is a pipeline that first summarizes a long, medical, user-written question, using a supervised summarization loss. Then, our system performs a two-step retrieval to return answers. The system first matches the summarized user question with an FAQ from a trusted medical knowledge base, and then retrieves a fixed number of relevant sentences from the corresponding answer document. In the absence of labels for question matching or answer relevance, we design 3 novel, self-supervised and semantically-guided losses. We evaluate our model against two strong retrieval-based question answering baselines. Evaluators ask their own questions and rate the answers retrieved by our baselines and own system according to their relevance. They find that our system retrieves more relevant answers, while achieving speeds 20 times faster. Our self-supervised losses also help the summarizer achieve higher scores in ROUGE, as well as in human evaluation metrics. We release our code to encourage further research.

We propose MM-Vet, an evaluation benchmark that examines large multimodal models (LMMs) on complicated multimodal tasks. Recent LMMs have shown various intriguing abilities, such as solving math problems written on the blackboard, reasoning about events and celebrities in news images, and explaining visual jokes. Rapid model advancements pose challenges to evaluation benchmark development. Problems include: (1) How to systematically structure and evaluate the complicated multimodal tasks; (2) How to design evaluation metrics that work well across question and answer types; and (3) How to give model insights beyond a simple performance ranking. To this end, we present MM-Vet, designed based on the insight that the intriguing ability to solve complicated tasks is often achieved by a generalist model being able to integrate different core vision-language (VL) capabilities. MM-Vet defines 6 core VL capabilities and examines the 16 integrations of interest derived from the capability combination. For evaluation metrics, we propose an LLM-based evaluator for open-ended outputs. The evaluator enables the evaluation across different question types and answer styles, resulting in a unified scoring metric. We evaluate representative LMMs on MM-Vet, providing insights into the capabilities of different LMM system paradigms and models. Code and data are available at https://github.com/yuweihao/MM-Vet.

Consistency is one of the major challenges faced by dialogue agents. A human-like dialogue agent should not only respond naturally, but also maintain a consistent persona. In this paper, we exploit the advantages of natural language inference (NLI) technique to address the issue of generating persona consistent dialogues. Different from existing work that re-ranks the retrieved responses through an NLI model, we cast the task as a reinforcement learning problem and propose to exploit the NLI signals from response-persona pairs as rewards for the process of dialogue generation. Specifically, our generator employs an attention-based encoder-decoder to generate persona-based responses. Our evaluator consists of two components: an adversarially trained naturalness module and an NLI based consistency module. Moreover, we use another well-performed NLI model in the evaluation of persona-consistency. Experimental results on both human and automatic metrics, including the model-based consistency evaluation, demonstrate that the proposed approach outperforms strong generative baselines, especially in the persona-consistency of generated responses.

We introduce DEsignBench, a text-to-image (T2I) generation benchmark tailored for visual design scenarios. Recent T2I models like DALL-E 3 and others, have demonstrated remarkable capabilities in generating photorealistic images that align closely with textual inputs. While the allure of creating visually captivating images is undeniable, our emphasis extends beyond mere aesthetic pleasure. We aim to investigate the potential of using these powerful models in authentic design contexts. In pursuit of this goal, we develop DEsignBench, which incorporates test samples designed to assess T2I models on both "design technical capability" and "design application scenario." Each of these two dimensions is supported by a diverse set of specific design categories. We explore DALL-E 3 together with other leading T2I models on DEsignBench, resulting in a comprehensive visual gallery for side-by-side comparisons. For DEsignBench benchmarking, we perform human evaluations on generated images in DEsignBench gallery, against the criteria of image-text alignment, visual aesthetic, and design creativity. Our evaluation also considers other specialized design capabilities, including text rendering, layout composition, color harmony, 3D design, and medium style. In addition to human evaluations, we introduce the first automatic image generation evaluator powered by GPT-4V. This evaluator provides ratings that align well with human judgments, while being easily replicable and cost-efficient. A high-resolution version is available at https://github.com/design-bench/design-bench.github.io/raw/main/designbench.pdf?download=

Existing LLM-based systems for writing long-form stories or story outlines frequently suffer from unnatural pacing, whether glossing over important events or over-elaborating on insignificant details, resulting in a jarring experience for the reader. We propose a CONCrete Outline ConTrol (CONCOCT) system to improve pacing when automatically generating story outlines. We first train a concreteness evaluator to judge which of two events is more concrete (low-level-detailed). This evaluator can then be used to control pacing in hierarchical outline generation; in this work, we explore a vaguest-first expansion procedure that aims for uniform pacing. We further use the evaluator to filter new outline items based on predicted concreteness. Compared to a baseline hierarchical outline generator, humans judge CONCOCT's pacing to be more consistent over 57% of the time across multiple outline lengths; the gains also translate to downstream stories. All code, data, and models are open-sourced.

Despite the remarkable coherence of Large Language Models (LLMs), existing evaluation methods often suffer from fluency bias and rely heavily on multiple-choice formats, making it difficult to assess factual accuracy and complex reasoning effectively. LLMs thus frequently generate factually inaccurate responses, especially in complex reasoning tasks, highlighting two prominent challenges: (1) the inadequacy of existing methods to evaluate reasoning and factual accuracy effectively, and (2) the reliance on human evaluators for nuanced judgment, as illustrated by Williams and Huckle (2024)[1], who found manual grading indispensable despite automated grading advancements. To address evaluation gaps in open-ended reasoning tasks, we introduce the EQUATOR Evaluator (Evaluation of Question Answering Thoroughness in Open-ended Reasoning). This framework combines deterministic scoring with a focus on factual accuracy and robust reasoning assessment. Using a vector database, EQUATOR pairs open-ended questions with human-evaluated answers, enabling more precise and scalable evaluations. In practice, EQUATOR significantly reduces reliance on human evaluators for scoring and improves scalability compared to Williams and Huckle's (2004)[1] methods. Our results demonstrate that this framework significantly outperforms traditional multiple-choice evaluations while maintaining high accuracy standards. Additionally, we introduce an automated evaluation process leveraging smaller, locally hosted LLMs. We used LLaMA 3.2B, running on the Ollama binaries to streamline our assessments. This work establishes a new paradigm for evaluating LLM performance, emphasizing factual accuracy and reasoning ability, and provides a robust methodological foundation for future research.

Chain-of-thought (CoT) reasoning greatly improves the interpretability and problem-solving abilities of multimodal large language models (MLLMs). However, existing approaches are focused on text CoT, limiting their ability to leverage visual cues. Visual CoT remains underexplored, and the only work is based on supervised fine-tuning (SFT) that relies on extensive labeled bounding-box data and is hard to generalize to unseen cases. In this paper, we introduce Unsupervised Visual CoT (UV-CoT), a novel framework for image-level CoT reasoning via preference optimization. UV-CoT performs preference comparisons between model-generated bounding boxes (one is preferred and the other is dis-preferred), eliminating the need for bounding-box annotations. We get such preference data by introducing an automatic data generation pipeline. Given an image, our target MLLM (e.g., LLaVA-1.5-7B) generates seed bounding boxes using a template prompt and then answers the question using each bounded region as input. An evaluator MLLM (e.g., OmniLLM-12B) ranks the responses, and these rankings serve as supervision to train the target MLLM with UV-CoT by minimizing negative log-likelihood losses. By emulating human perception--identifying key regions and reasoning based on them--UV-CoT can improve visual comprehension, particularly in spatial reasoning tasks where textual descriptions alone fall short. Our experiments on six datasets demonstrate the superiority of UV-CoT, compared to the state-of-the-art textual and visual CoT methods. Our zero-shot testing on four unseen datasets shows the strong generalization of UV-CoT. The code is available in https://github.com/kesenzhao/UV-CoT.

Interpreting the law is always essential for the law to adapt to the ever-changing society. It is a critical and challenging task even for legal practitioners, as it requires meticulous and professional annotations and summarizations by legal experts, which are admittedly time-consuming and expensive to collect at scale. To alleviate the burden on legal experts, we propose a method for automated legal interpretation. Specifically, by emulating doctrinal legal research, we introduce a novel framework, ATRIE, to address Legal Concept Interpretation, a typical task in legal interpretation. ATRIE utilizes large language models (LLMs) to AuTomatically Retrieve concept-related information, Interpret legal concepts, and Evaluate generated interpretations, eliminating dependence on legal experts. ATRIE comprises a legal concept interpreter and a legal concept interpretation evaluator. The interpreter uses LLMs to retrieve relevant information from previous cases and interpret legal concepts. The evaluator uses performance changes on Legal Concept Entailment, a downstream task we propose, as a proxy of interpretation quality. Automated and multifaceted human evaluations indicate that the quality of our interpretations is comparable to those written by legal experts, with superior comprehensiveness and readability. Although there remains a slight gap in accuracy, it can already assist legal practitioners in improving the efficiency of legal interpretation.

Recent advances in motion-aware large language models have shown remarkable promise for unifying motion understanding and generation tasks. However, these models typically treat understanding and generation separately, limiting the mutual benefits that could arise from interactive feedback between tasks. In this work, we reveal that motion assessment and refinement tasks act as crucial bridges to enable bidirectional knowledge flow between understanding and generation. Leveraging this insight, we propose Interleaved Reasoning for Motion Generation (IRMoGen), a novel paradigm that tightly couples motion generation with assessment and refinement through iterative text-motion dialogue. To realize this, we introduce IRG-MotionLLM, the first model that seamlessly interleaves motion generation, assessment, and refinement to improve generation performance. IRG-MotionLLM is developed progressively with a novel three-stage training scheme, initializing and subsequently enhancing native IRMoGen capabilities. To facilitate this development, we construct an automated data engine to synthesize interleaved reasoning annotations from existing text-motion datasets. Extensive experiments demonstrate that: (i) Assessment and refinement tasks significantly improve text-motion alignment; (ii) Interleaving motion generation, assessment, and refinement steps yields consistent performance gains across training stages; and (iii) IRG-MotionLLM clearly outperforms the baseline model and achieves advanced performance on standard text-to-motion generation benchmarks. Cross-evaluator testing further validates its effectiveness. Code & Data: https://github.com/HumanMLLM/IRG-MotionLLM/tree/main.

To enhance the controllability of text-to-image diffusion models, current ControlNet-like models have explored various control signals to dictate image attributes. However, existing methods either handle conditions inefficiently or use a fixed number of conditions, which does not fully address the complexity of multiple conditions and their potential conflicts. This underscores the need for innovative approaches to manage multiple conditions effectively for more reliable and detailed image synthesis. To address this issue, we propose a novel framework, DynamicControl, which supports dynamic combinations of diverse control signals, allowing adaptive selection of different numbers and types of conditions. Our approach begins with a double-cycle controller that generates an initial real score sorting for all input conditions by leveraging pre-trained conditional generation models and discriminative models. This controller evaluates the similarity between extracted conditions and input conditions, as well as the pixel-level similarity with the source image. Then, we integrate a Multimodal Large Language Model (MLLM) to build an efficient condition evaluator. This evaluator optimizes the ordering of conditions based on the double-cycle controller's score ranking. Our method jointly optimizes MLLMs and diffusion models, utilizing MLLMs' reasoning capabilities to facilitate multi-condition text-to-image (T2I) tasks. The final sorted conditions are fed into a parallel multi-control adapter, which learns feature maps from dynamic visual conditions and integrates them to modulate ControlNet, thereby enhancing control over generated images. Through both quantitative and qualitative comparisons, DynamicControl demonstrates its superiority over existing methods in terms of controllability, generation quality and composability under various conditional controls.

Interpretation is essential to deciphering the language of art: audiences communicate with artists by recovering meaning from visual artifacts. However, current Generative Art (GenArt) evaluators remain fixated on surface-level image quality or literal prompt adherence, failing to assess the deeper symbolic or abstract meaning intended by the creator. We address this gap by formalizing a Peircean computational semiotic theory that models Human-GenArt Interaction (HGI) as cascaded semiosis. This framework reveals that artistic meaning is conveyed through three modes - iconic, symbolic, and indexical - yet existing evaluators operate heavily within the iconic mode, remaining structurally blind to the latter two. To overcome this structural blindness, we propose SemJudge. This evaluator explicitly assesses symbolic and indexical meaning in HGI via a Hierarchical Semiosis Graph (HSG) that reconstructs the meaning-making process from prompt to generated artifact. Extensive quantitative experiments show that SemJudge aligns more closely with human judgments than prior evaluators on an interpretation-intensive fine-art benchmark. User studies further demonstrate that SemJudge produces deeper, more insightful artistic interpretations, thereby paving the way for GenArt to move beyond the generation of "pretty" images toward a medium capable of expressing complex human experience. Project page: https://github.com/songrise/SemJudge.

Evaluating LLM mediators remains challenging, as mediation unfolds as a real-time trajectory shaped by disputants' shifting emotions, intentions, and context. Existing testbeds rely on a few expert-authored domains, vary mainly strategic posture, and score every turn against every topic, introducing off-topic noise. We introduce SoCRATES, a benchmark for evaluating proactive LLM mediators in realistic, multi-domain testbeds. It constructs scenarios from real conflicts through an agentic pipeline across eight domains, probes five socio-cognitive adaptation axes (strategic posture, party composition, history length, emotional reactivity, and cultural identity), and scores each topic only on the turns that advance it via a topic-localized evaluator. The evaluator reaches 0.82 alignment with human experts, more than doubling a per-turn baseline. Benchmarking eight frontier LLMs, we find that even the strongest mediator closes only about a third of the unmediated consensus gap under diverse and realistic testbeds, with performance varying sharply by socio-cognitive axis, highlighting that progress lies in social adaptation to diverse conditions.

Modern machine learning is still largely organized around a single recipe: choose a parameterized model family and optimize its weights. Although highly successful, this paradigm is too narrow for many structured prediction problems, where the main bottleneck is not parameter fitting but discovering what should be computed from the data. Success often depends on identifying the right transformations, statistics, invariances, interaction structures, temporal summaries, gates, or nonlinear compositions, especially when objectives are non-differentiable, evaluation is cross-validation-based, interpretability matters, or continual adaptation is required. We present EvoForest, a hybrid neuro-symbolic system for end-to-end open-ended evolution of computation. Rather than merely generating features, EvoForest jointly evolves reusable computational structure, callable function families, and trainable low-dimensional continuous components inside a shared directed acyclic graph. Intermediate nodes store alternative implementations, callable nodes encode reusable transformation families such as projections, gates, and activations, output nodes define candidate predictive computations, and persistent global parameters can be refined by gradient descent. For each graph configuration, EvoForest evaluates the discovered computation and uses a lightweight Ridge-based readout to score the resulting representation against a non-differentiable cross-validation target. The evaluator also produces structured feedback that guides future LLM-driven mutations. In the 2025 ADIA Lab Structural Break Challenge, EvoForest reached 94.13% ROC-AUC after 600 evolution steps, exceeding the publicly reported winning score of 90.14% under the same evaluation protocol.

Pre-Phase-A design of lunar micro-rovers is dominated by tightly coupled mobility, power, thermal, and mass trades, yet conceptual-design tooling for the rapidly growing sub-50 kg class is typically proprietary, weakly benchmarked, or too slow to drive optimization. We contribute RoverDevKit, an open analytical evaluator coupling terramechanics, mass, power, thermal survival, and traverse that runs in 30ms per mission, fast enough to serve directly as a multi-objective optimizer's fitness function. Across mare, polar, highland, and crater-rim scenarios, NSGA-II Pareto fronts show that the binding design trade changes with mission profile within a single mass class: energy storage dominates at high latitude, slope traction on loose highland regolith, and traverse range on mare and crater-rim missions. Notably, rigid four-wheel layouts Pareto-dominate the full modeled mass range under smooth-regolith range-mass-slope objectives, contrary to the expectation that six-wheel architectures become optimal at heavier masses; six-wheel rocker-bogie layouts enter the Pareto set only once missions impose an obstacle-navigation requirement. The evaluator performance is benchmarked using both component and system checks: the terramechanics kernel matches measured single-wheel drawbar pull within the literature model-form band on two independent datasets, the bottom-up mass model predicts published in-class (5-50 kg) rover masses to 13.3% median absolute error, and a rediscovery check places real micro-rovers near the optimizer's fronts. Propagating the measured terramechanics error through the optimizer leaves the qualitative design rules unchanged. The tool, data, validation artifacts, and figure-generation scripts are released openly.

Video world models can produce visually convincing yet physically inconsistent sequences, raising concerns about their reliability for prediction and planning in embodied AI systems. Existing evaluations often rely on model-based judgments or reference videos, while direct physical tests largely focus on mechanics. We introduce World Models' Last Exam in Physics, a measurement-based benchmark for evaluating physical consistency in video world models. The benchmark comprises 40 controlled tasks spanning mechanics, optics, fluids, thermal and phase-change phenomena, electromagnetism, and surface tension. Each task pairs an initial image and a generation prompt with predefined physical criteria, enabling interpretable tests of observable physical relationships without requiring reference videos. Its evaluator combines task-observability screening with task-specific quantitative physical measurements. Experiments on eight video generation models across 1,280 videos reveal persistent physical inconsistencies and substantial variation across tasks, with the best model achieving an overall score of 57.76 out of 100. Evaluation on synthetic videos with known physical relationships provides evidence for the validity of the measurement module under controlled conditions. The evaluator also achieves higher agreement with human judgments than a direct vision-language model baseline in both within-task rankings and pairwise comparisons. By combining coverage across physical domains with scores grounded in measurable evidence and explicit measurement limitations, the benchmark provides an interpretable basis for diagnosing physical inconsistencies and tracking progress toward physically consistent video world models.

Developing agents for hardware design and verification requires reliable correctness feedback. As a hardware specification may permit correct implementations with different latencies, matching design and reference outputs cycle by cycle can reject valid designs. To address this, we introduce BEHAVE, an agentic framework for multi-turn joint hardware design and verification through functional behavior modeling. We define Behavior IR to express task functionality as executable behavior models without prescribing implementation timing beyond the specification. The agent iteratively develops a register-transfer-level (RTL) design and a behavior model as the design's verification reference. Our evaluator, BEHAVE-Sim, checks both artifacts separately against a hidden golden behavior model using input stimuli generated by random sampling and solver-guided search. BEHAVE thus supports power, performance, and area (PPA) exploration across task-permitted latencies and microarchitectures. During training, the same evaluator provides verifiable reinforcement learning (RL) rewards from specification-behavior pairs without reference RTL. For self-improvement, the agent continually searches for high-level implementations relevant to its capability gaps, constructs and checks specification-behavior pairs, and trains on the expanded task pool. We release BEHAVE-Train and BEHAVE-Eval with 600 human-reviewed specification-behavior pairs for realistic hardware workloads. Starting from 60 seed tasks and acquiring 100 new tasks, self-improvement raises Qwen3.8-27B's RTL pass@1 on BEHAVE-Eval from 55.0% to 75.0%, reaching performance comparable to RL using a 540-task pool.

Question generation (QG) is a natural language generation task where a model is trained to ask questions corresponding to some input text. Most recent approaches frame QG as a sequence-to-sequence problem and rely on additional features and mechanisms to increase performance; however, these often increase model complexity, and can rely on auxiliary data unavailable in practical use. A single Transformer-based unidirectional language model leveraging transfer learning can be used to produce high quality questions while disposing of additional task-specific complexity. Our QG model, finetuned from GPT-2 Small, outperforms several paragraph-level QG baselines on the SQuAD dataset by 0.95 METEOR points. Human evaluators rated questions as easy to answer, relevant to their context paragraph, and corresponding well to natural human speech. Also introduced is a new set of baseline scores on the RACE dataset, which has not previously been used for QG tasks. Further experimentation with varying model capacities and datasets with non-identification type questions is recommended in order to further verify the robustness of pretrained Transformer-based LMs as question generators.

Building open-domain conversational systems (or chatbots) that produce convincing responses is a recognized challenge. Recent state-of-the-art (SoTA) transformer-based models for the generation of natural language dialogue have demonstrated impressive performance in simulating human-like, single-turn conversations in English. This work investigates, by an empirical study, the potential for transfer learning of such models to Swedish language. DialoGPT, an English language pre-trained model, is adapted by training on three different Swedish language conversational datasets obtained from publicly available sources. Perplexity score (an automated intrinsic language model metric) and surveys by human evaluation were used to assess the performances of the fine-tuned models, with results that indicate that the capacity for transfer learning can be exploited with considerable success. Human evaluators asked to score the simulated dialogue judged over 57% of the chatbot responses to be human-like for the model trained on the largest (Swedish) dataset. We provide the demos and model checkpoints of our English and Swedish chatbots on the HuggingFace platform for public use.

We introduce anticipation: a method for constructing a controllable generative model of a temporal point process (the event process) conditioned asynchronously on realizations of a second, correlated process (the control process). We achieve this by interleaving sequences of events and controls, such that controls appear following stopping times in the event sequence. This work is motivated by problems arising in the control of symbolic music generation. We focus on infilling control tasks, whereby the controls are a subset of the events themselves, and conditional generation completes a sequence of events given the fixed control events. We train anticipatory infilling models using the large and diverse Lakh MIDI music dataset. These models match the performance of autoregressive models for prompted music generation, with the additional capability to perform infilling control tasks, including accompaniment. Human evaluators report that an anticipatory model produces accompaniments with similar musicality to even music composed by humans over a 20-second clip.

In multiplayer, first-person shooter games like Counter-Strike: Global Offensive (CS:GO), coordinated movement is a critical component of high-level strategic play. However, the complexity of team coordination and the variety of conditions present in popular game maps make it impractical to author hand-crafted movement policies for every scenario. We show that it is possible to take a data-driven approach to creating human-like movement controllers for CS:GO. We curate a team movement dataset comprising 123 hours of professional game play traces, and use this dataset to train a transformer-based movement model that generates human-like team movement for all players in a "Retakes" round of the game. Importantly, the movement prediction model is efficient. Performing inference for all players takes less than 0.5 ms per game step (amortized cost) on a single CPU core, making it plausible for use in commercial games today. Human evaluators assess that our model behaves more like humans than both commercially-available bots and procedural movement controllers scripted by experts (16% to 59% higher by TrueSkill rating of "human-like"). Using experiments involving in-game bot vs. bot self-play, we demonstrate that our model performs simple forms of teamwork, makes fewer common movement mistakes, and yields movement distributions, player lifetimes, and kill locations similar to those observed in professional CS:GO match play.

Extensive evaluation on a large number of word embedding models for language processing applications is conducted in this work. First, we introduce popular word embedding models and discuss desired properties of word models and evaluation methods (or evaluators). Then, we categorize evaluators into intrinsic and extrinsic two types. Intrinsic evaluators test the quality of a representation independent of specific natural language processing tasks while extrinsic evaluators use word embeddings as input features to a downstream task and measure changes in performance metrics specific to that task. We report experimental results of intrinsic and extrinsic evaluators on six word embedding models. It is shown that different evaluators focus on different aspects of word models, and some are more correlated with natural language processing tasks. Finally, we adopt correlation analysis to study performance consistency of extrinsic and intrinsic evalutors.

LLM-as-a-Judge and reward models are widely used alternatives of multiple-choice questions or human annotators for large language model (LLM) evaluation. Their efficacy shines in evaluating long-form responses, serving a critical role as evaluators of leaderboards and as proxies to align LLMs via reinforcement learning. However, despite their popularity, their effectiveness outside of English remains largely unexplored. In this paper, we conduct a comprehensive analysis on automated evaluators, reporting key findings on their behavior in a non-English environment. First, we discover that English evaluation capabilities significantly influence language-specific capabilities, often more than the language proficiency itself, enabling evaluators trained in English to easily transfer their skills to other languages. Second, we identify critical shortcomings, where LLMs fail to detect and penalize errors, such as factual inaccuracies, cultural misrepresentations, and the presence of unwanted language. Finally, we release Kudge, the first non-English meta-evaluation dataset containing 5,012 human annotations in Korean.

Understanding sources of a model's uncertainty regarding its predictions is crucial for effective human-AI collaboration. Prior work proposes using numerical uncertainty or hedges ("I'm not sure, but ..."), which do not explain uncertainty that arises from conflicting evidence, leaving users unable to resolve disagreements or rely on the output. We introduce CLUE (Conflict-and-Agreement-aware Language-model Uncertainty Explanations), the first framework to generate natural language explanations of model uncertainty by (i) identifying relationships between spans of text that expose claim-evidence or inter-evidence conflicts and agreements that drive the model's predictive uncertainty in an unsupervised way, and (ii) generating explanations via prompting and attention steering that verbalize these critical interactions. Across three language models and two fact-checking datasets, we show that CLUE produces explanations that are more faithful to the model's uncertainty and more consistent with fact-checking decisions than prompting for uncertainty explanations without span-interaction guidance. Human evaluators judge our explanations to be more helpful, more informative, less redundant, and more logically consistent with the input than this baseline. CLUE requires no fine-tuning or architectural changes, making it plug-and-play for any white-box language model. By explicitly linking uncertainty to evidence conflicts, it offers practical support for fact-checking and generalises readily to other tasks that require reasoning over complex information.

Reliable evaluation of large language models (LLMs) is impeded by two key challenges: objective metrics often fail to reflect human perception of natural language, and exhaustive human labeling is prohibitively expensive. Here, we propose a sample-efficient human evaluation method for LLMs based on the principle of MAximum Discrepancy (MAD) Competition. Our method automatically and adaptively selects a compact set of input instructions that maximize semantic discrepancy between pairs of LLM responses. Human evaluators then perform three-alternative forced choices on these paired responses, which are aggregated into a global ranking using Elo rating. We apply our approach to compare eight widely used LLMs across four tasks: scientific knowledge understanding, mathematical reasoning, creative and functional writing, and code generation and explanation. Experimental results show that our sample-efficient evaluation method recovers "gold-standard" model rankings with a handful of MAD-selected instructions, reveals respective strengths and weaknesses of each LLM, and offers nuanced insights to guide future LLM development. Code is available at https://github.com/weiji-Feng/MAD-Eval .

We explore the automatic generation of interactive, scenario-based lessons designed to train novice human tutors who teach middle school mathematics online. Employing prompt engineering through a Retrieval-Augmented Generation approach with GPT-4o, we developed a system capable of creating structured tutor training lessons. Our study generated lessons in English for three key topics: Encouraging Students' Independence, Encouraging Help-Seeking Behavior, and Turning on Cameras, using a task decomposition prompting strategy that breaks lesson generation into sub-tasks. The generated lessons were evaluated by two human evaluators, who provided both quantitative and qualitative evaluations using a comprehensive rubric informed by lesson design research. Results demonstrate that the task decomposition strategy led to higher-rated lessons compared to single-step generation. Human evaluators identified several strengths in the LLM-generated lessons, including well-structured content and time-saving potential, while also noting limitations such as generic feedback and a lack of clarity in some instructional sections. These findings underscore the potential of hybrid human-AI approaches for generating effective lessons in tutor training.

As large language models (LLMs) become increasingly integrated into society, their alignment with human morals is crucial. To better understand this alignment, we created a large corpus of human- and LLM-generated responses to various moral scenarios. We found a misalignment between human and LLM moral assessments; although both LLMs and humans tended to reject morally complex utilitarian dilemmas, LLMs were more sensitive to personal framing. We then conducted a quantitative user study involving 230 participants (N=230), who evaluated these responses by determining whether they were AI-generated and assessed their agreement with the responses. Human evaluators preferred LLMs' assessments in moral scenarios, though a systematic anti-AI bias was observed: participants were less likely to agree with judgments they believed to be machine-generated. Statistical and NLP-based analyses revealed subtle linguistic differences in responses, influencing detection and agreement. Overall, our findings highlight the complexities of human-AI perception in morally charged decision-making.

Recent advances in text-to-video generation have produced increasingly realistic and diverse content, yet evaluating such videos remains a fundamental challenge due to their multi-faceted nature encompassing visual quality, semantic alignment, and physical consistency. Existing evaluators and reward models are limited to single opaque scores, lack interpretability, or provide only coarse analysis, making them insufficient for capturing the comprehensive nature of video quality assessment. We present VideoScore2, a multi-dimensional, interpretable, and human-aligned framework that explicitly evaluates visual quality, text-to-video alignment, and physical/common-sense consistency while producing detailed chain-of-thought rationales. Our model is trained on a large-scale dataset VideoFeedback2 containing 27,168 human-annotated videos with both scores and reasoning traces across three dimensions, using a two-stage pipeline of supervised fine-tuning followed by reinforcement learning with Group Relative Policy Optimization (GRPO) to enhance analytical robustness. Extensive experiments demonstrate that VideoScore2 achieves superior performance with 44.35 (+5.94) accuracy on our in-domain benchmark VideoScore-Bench-v2 and 50.37 (+4.32) average performance across four out-of-domain benchmarks (VideoGenReward-Bench, VideoPhy2, etc), while providing interpretable assessments that bridge the gap between evaluation and controllable generation through effective reward modeling for Best-of-N sampling. Project Page: https://tiger-ai-lab.github.io/VideoScore2/

Despite rapid advances in text-to-video synthesis, generated video quality remains critically dependent on precise user prompts. Existing test-time optimization methods, successful in other domains, struggle with the multi-faceted nature of video. In this work, we introduce VISTA (Video Iterative Self-improvemenT Agent), a novel multi-agent system that autonomously improves video generation through refining prompts in an iterative loop. VISTA first decomposes a user idea into a structured temporal plan. After generation, the best video is identified through a robust pairwise tournament. This winning video is then critiqued by a trio of specialized agents focusing on visual, audio, and contextual fidelity. Finally, a reasoning agent synthesizes this feedback to introspectively rewrite and enhance the prompt for the next generation cycle. Experiments on single- and multi-scene video generation scenarios show that while prior methods yield inconsistent gains, VISTA consistently improves video quality and alignment with user intent, achieving up to 60% pairwise win rate against state-of-the-art baselines. Human evaluators concur, preferring VISTA outputs in 66.4% of comparisons.

While current Computer Use Agent (CUA) benchmarks measure task completion effectively, they provide limited assessment of enterprise deployment readiness, emphasizing functional correctness over the operational reliability required for production systems. We present UI-CUBE (UiPath Computer Use BEnchmark), a systematic benchmark comprising 226 tasks across two difficulty tiers designed to expose fundamental architectural limitations in current CUAs. Our evaluation covers simple UI interactions (136 tasks) and complex workflows including copy-paste tasks (50 tasks) and enterprise application scenarios (40 tasks), with systematic interface variation coverage, multi-resolution testing and automated validation of task success through the application state. Evaluation of five state-of-the-art models reveals a sharp capability cliff rather than gradual performance degradation. Simple UI interactions achieve 67-85% success rates (compared to 97.9% human performance), but complex workflows drop precipitously to 9-19%. Human evaluators with no prior application experience achieve only 61.2% on complex tasks despite near-perfect performance on simple tasks, establishing realistic performance ceilings. This discontinuous performance pattern -- where agents achieve 68-87% of human performance on simple tasks but only 15-32% on complex workflows -- indicates fundamental architectural limitations in memory management, hierarchical planning, and state coordination rather than incremental capability gaps addressable through better training or prompting. UI-CUBE functions as an enterprise-readiness diagnostic, revealing that while current CUAs can manipulate individual interface elements, they cannot yet function as reliable workflow automation tools. These findings provide architectural insights essential for developing production-ready CUAs capable of managing complex, multi-step enterprise processes.

People are remarkably capable of generating their own goals, beginning with child's play and continuing into adulthood. Despite considerable empirical and computational work on goals and goal-oriented behavior, models are still far from capturing the richness of everyday human goals. Here, we bridge this gap by collecting a dataset of human-generated playful goals (in the form of scorable, single-player games), modeling them as reward-producing programs, and generating novel human-like goals through program synthesis. Reward-producing programs capture the rich semantics of goals through symbolic operations that compose, add temporal constraints, and allow for program execution on behavioral traces to evaluate progress. To build a generative model of goals, we learn a fitness function over the infinite set of possible goal programs and sample novel goals with a quality-diversity algorithm. Human evaluators found that model-generated goals, when sampled from partitions of program space occupied by human examples, were indistinguishable from human-created games. We also discovered that our model's internal fitness scores predict games that are evaluated as more fun to play and more human-like.

Online Reinforcement Learning (RL) offers a promising avenue for complex image editing but is currently constrained by the scarcity of reliable and fine-grained reward signals. Existing evaluators frequently struggle with a critical perception gap we term "Attention Collapse," where models neglect cross-image comparisons and fail to capture fine-grained details, resulting in inaccurate perception and miscalibrated scores. To address these limitations, we propose SpatialReward, a reward model that enforces precise verification via explicit spatial reasoning. By anchoring reasoning to predicted edit regions, SpatialReward grounds semantic judgments in pixel-level evidence, significantly enhancing evaluative accuracy. Trained on a curated 260k spatial-aware dataset, our model achieves state-of-the-art performance on MMRB2 and EditReward-Bench, and outperforms proprietary evaluators on our proposed MultiEditReward-Bench. Furthermore, SpatialReward serves as a robust signal in online RL, boosting OmniGen2 by +0.90 on GEdit-Bench--surpassing the leading discriminative model and doubling the gain of GPT-4.1 (+0.45). These results demonstrate that spatial reasoning is essential for unlocking effective alignment in image editing.

Automatic evaluators such as reward models play a central role in the alignment and evaluation of large vision-language models (LVLMs). Despite their growing importance, these evaluators are almost exclusively assessed on English-centric benchmarks, leaving open the question of how well these evaluators generalize across languages. To answer this question, we introduce MM-JudgeBench, the first large-scale benchmark for multilingual and multimodal judge model evaluation, which includes over 60K pairwise preference instances spanning 25 typologically diverse languages. MM-JudgeBench integrates two complementary subsets: a general vision-language preference evaluation subset extending VL-RewardBench, and a chart-centric visual-text reasoning subset derived from OpenCQA, enabling systematic analysis of reward models (i.e., LVLM judges) across diverse settings. We additionally release a multilingual training set derived from MM-RewardBench, disjoint from our evaluation data, to support domain adaptation. By evaluating 22 LVLMs (15 open-source, 7 proprietary), we uncover substantial cross-lingual performance variance in our proposed benchmark. Our analysis further shows that model size and architecture are poor predictors of multilingual robustness, and that even state-of-the-art LVLM judges exhibit inconsistent behavior across languages. Together, these findings expose fundamental limitations of current reward modeling and underscore the necessity of multilingual, multimodal benchmarks for developing reliable automated evaluators.

In this report, we introduce Qwen-Music, a powerful music generation model capable of producing highly musical and high-fidelity songs with complete vocal singing. Qwen-Music supports two core tasks: Text to Music Generation, which create entirely new songs from text descriptions, lyrics, and musical attributes, and Cover Song Generation, which reinterprets existing songs with different styles and vocal characteristics. Architecturally, Qwen-Music integrates three core components: Qwen-Music-Tokenizer, Qwen-Music-LLM, and Qwen-Music-Render. Qwen-Music-Tokenizer compresses audio into a 25 Hz single-codebook stream of Music Semantic Tokens that preserve semantic and melodic information for LLM prediction. Based on these tokens, Qwen-Music-LLM performs autoregressive music semantic modeling, with a key novelty being a melody-token-based chain-of-thought (Melody-CoT) mechanism that plans melodies before full-song generation, improving creativity, musicality, structural coherence, and reference-audio-based melody cloning. To overcome the fidelity limitations of discrete semantic tokens, Qwen-Music-Render performs generative stereo rendering, enriching acoustic details and producing high-fidelity stereo waveforms. Finally, we train Qwen-Music-LLM on more than 5 million hours of multilingual music data covering hundreds of languages. We first apply quality-aware pre-training curriculum, then use progressive post-training, comprising supervised initialization, offline DPO, and online GSPO, to further improve musicality and instruction-following ability. Across 600 Chinese and English prompts, Qwen-Music achieves state-of-the-art results in 13 of 16 objective musicality and audio-quality metrics. Professional evaluators also prefer Qwen-Music over leading proprietary systems. For cover song generation, Qwen-Music preserves reference melodies more accurately than leading proprietary systems.

Executable evaluation -- checking the consequences of an agent's actions with a program rather than grading its prose -- has become a prominent way to assess tool-using AI agents in software settings. Electric power engineering has not yet had an analogous benchmark: language-model use is still dominated by retrieval and text question answering, while agents acting on power-system artifacts remain mostly academic prototypes. We introduce the Power Systems Agent Benchmark, an executable benchmark for power-engineering agents. An agent receives a structured task and returns a structured solution; a deterministic evaluator recomputes the engineering quantities, checks operational constraints, and returns a feasibility flag, a normalized score, and explicit violations. The benchmark contains 41 task families across eight areas of power engineering, from power flow and protection to stability, microgrids, reliability, power quality, and forecasting. Each task is grounded in a citable source, standard, or documented engineering formulation. To resist contamination, held-out cases are synthesized on demand by per-family generators from private seeds: the construction is inspectable, but the instances remain private. In a reference evaluation with three command-line agents, the strongest score near the compact tier's ceiling, a smaller open model trails, and public and held-out performance are broadly consistent; a separate public-split grid with OpenCode and Aider probes harness effects. The reference evaluation doubles as quality control: unanimous failures flag candidate task or evaluator defects, and it exposed a latent evaluator bug missed by self-consistency checks. The evaluators are compact deterministic surrogates, but the task contract allows their internals to be upgraded to simulator-backed checks without changing how tasks are posed or solved.

In many social-science research tasks, such as economics, LLM-based agents must produce outputs for which no cheap, task-complete, machine-readable correctness signal exists. This creates a distinctive reliability problem for multi-agent systems: how should generation, critique, coordination, and human judgment be organized when no component can certify the final result? We address this problem through pAI-Econ-claude, a gated, human-in-the-loop multi-agent architecture for AI-assisted economic theory development. Agents coordinate through a shared workspace of inspectable intermediate records; specialized gates diagnose targeted failure modes and recommend loopbacks without certifying correctness; and human checkpoints retain authority over decisions that are costly to reverse. We evaluate the architecture on five matched economic-theory tasks against an ungated baseline. Two evaluators blinded to configuration agreed on all five pairwise rankings, preferring the gated architecture in four tasks and the baseline in one. Mean failure severity fell from 1.58 to 1.16, while overall usefulness rose from 2.60 to 3.10. The largest observed gain occurred when a reality check rejected a false market-structure premise and a proof review prompted revision of a false welfare claim. The negative case shows that scaffolding can also compress an economically important mechanism too aggressively. The results support a bounded claim: gated oversight improves the auditability of AI-assisted economic theory without substituting for formal verification, and the allocation of irreversible human judgment is a more informative design variable than pure agent autonomy. The workflow is publicly available at https://github.com/maxwell2732/pAI-Econ-claude.

An AI scientist should not grade its own homework. Yet in the systems we examined, the agent that reviews the work usually comes from the same model family as the agent that produced it, or at least from the same vendor. Model evaluators are known to favour their own generations. Whether models trained alike also share blind spots is a conjecture, not a settled finding, but if they do, the reviewer inherits the author's. The record of what was flagged and what was waved through often sits in platform logs that nobody outside can replay. We present CrossAudit, a protocol for supervising autonomous research pipelines. It rests on three commitments. Each increment of work is audited by an agent from a different vendor against a rulebook a human wrote and versioned. Reports, verdicts, disputes and rulings are git commits, so the supervision history can be re-read and cited; raw model exchanges are not yet part of that record. Scripted checks run before any model does. Advisory judgement never gates the pipeline: a model blocks only by citing a rule, and no model may waive a deterministic failure. Blockers that survive a bounded number of revision rounds go to a person. We state the protocol as eight invariants. We describe a reference implementation built from GitHub Actions and a few hundred lines of Python, and report a live deployment of a closely related variant in a computational-chemistry pipeline. We also ran a seeded-defect trial (30 increments, 43 seeded defects, one run per configuration). A cross-vendor audit of our own repository then voided its blinding. We adopt that audit's findings and report the corrected results. The trial shows that two vendors read the same rulebook differently. It does not show that either is better. The strongest evidence here is the committed, uncontrolled record of cross-vendor audits of this paper itself.

Training language models for formal theorem proving requires large numbers of verifier calls. Slow checks and timeouts often leave GPUs waiting and withhold rewards from potentially correct proofs. We train neural evaluators to predict Lean 4 verdicts and compiler feedback on GPUs using 4.84M labelled programs. We compare two validity classifiers with a feedback model that annotates proof attempts with predicted goal states and error messages. A two-stage evaluator that screens candidates with a classifier and scores its positive predictions with the feedback model reaches 87.8% accuracy in predicting whether Lean accepts a candidate proof, evaluated on a 68,932-program benchmark collected from public provers. The combined evaluator processes 3.5 candidates per second on one H100, 4.2x the throughput of our 64-worker Lean baseline with environment reuse. Finally, we combine our evaluator with Lean verdicts to obtain a hybrid reward system that shortens GRPO training steps by approximately 31%. After one GRPO epoch with hybrid rewards, the prover achieves nearly the same miniF2F pass@1 as with Lean-only rewards (62.2% versus 62.4%) and a pass@32 of 72.4% versus 71.3%.

Long-horizon tool agents often make useful progress without reaching terminal success, motivating partial-credit evaluation. Yet evaluators may reward milestones that were temporary, later reversed, or not attributable to the evaluated agent. Comparing an honest trajectory with a higher-scoring adversarial one is inconclusive if the latter made more genuine progress. We introduce PartHackBench, a controlled methodology that removes this confound. A private certifier admits a pair only when its trajectories match component-wise in both current-state predicate satisfaction and standardized agent attribution; score inflation, defined as f(A) - f(H), is measured only afterward. In 18 sealed held-out tasks in PB-CSTE, the frozen historical-target run produced matched adversaries for 15 tasks. Historical credit yielded mean inflation of .252, conditional attack success of 10/15, end-to-end yield of 10/18, and detected none of 14 strict rollbacks. Semantic LLM judges were more resistant but remained vulnerable, especially under evaluator-targeted attacks, while PB-CSTE current-state controls, defined as exact functions of the certified components, yielded zero inflation by construction. PartHackBench thus provides a certified control for testing whether evaluator credit changes while all benchmark-defined task-relevant progress remains fixed.

Evaluating the quality and variability of text generated by Large Language Models (LLMs) poses a significant, yet unresolved research challenge. Traditional evaluation methods, such as ROUGE and BERTScore, which measure token similarity, often fail to capture the holistic semantic equivalence. This results in a low correlation with human judgments and intuition, which is especially problematic in high-stakes applications like healthcare and finance where reliability, safety, and robust decision-making are highly critical. This work proposes DCR, an automated framework for evaluating and improving the consistency of LLM-generated texts using a divide-conquer-reasoning approach. Unlike existing LLM-based evaluators that operate at the paragraph level, our method employs a divide-and-conquer evaluator (DCE) that breaks down the paragraph-to-paragraph comparison between two generated responses into individual sentence-to-paragraph comparisons, each evaluated based on predefined criteria. To facilitate this approach, we introduce an automatic metric converter (AMC) that translates the output from DCE into an interpretable numeric score. Beyond the consistency evaluation, we further present a reason-assisted improver (RAI) that leverages the analytical reasons with explanations identified by DCE to generate new responses aimed at reducing these inconsistencies. Through comprehensive and systematic empirical analysis, we show that our approach outperforms state-of-the-art methods by a large margin (e.g., +19.3% and +24.3% on the SummEval dataset) in evaluating the consistency of LLM generation across multiple benchmarks in semantic, factual, and summarization consistency tasks. Our approach also substantially reduces nearly 90% of output inconsistencies, showing promise for effective hallucination mitigation.

Large Language Models (LLMs) have exhibited remarkable success in long-form context comprehension tasks. However, their capacity to generate long contents, such as reports and articles, remains insufficiently explored. Current benchmarks do not adequately assess LLMs' ability to produce informative and comprehensive content, necessitating a more rigorous evaluation approach. In this study, we introduce ProxyQA, a framework for evaluating long-form text generation, comprising in-depth human-curated meta-questions spanning various domains. Each meta-question contains corresponding proxy-questions with annotated answers. LLMs are prompted to generate extensive content in response to these meta-questions. Utilizing an evaluator and incorporating generated content as background context, ProxyQA evaluates the quality of generated content based on the evaluator's performance in answering the proxy-questions. We examine multiple LLMs, emphasizing ProxyQA's demanding nature as a high-quality assessment tool. Human evaluation demonstrates that evaluating through proxy-questions is a highly self-consistent and human-criteria-correlated validation method. The dataset and leaderboard will be available at https://github.com/Namco0816/ProxyQA.

In the rapidly evolving landscape of Large Language Models (LLMs), ensuring robust safety measures is paramount. To meet this crucial need, we propose SALAD-Bench, a safety benchmark specifically designed for evaluating LLMs, attack, and defense methods. Distinguished by its breadth, SALAD-Bench transcends conventional benchmarks through its large scale, rich diversity, intricate taxonomy spanning three levels, and versatile functionalities.SALAD-Bench is crafted with a meticulous array of questions, from standard queries to complex ones enriched with attack, defense modifications and multiple-choice. To effectively manage the inherent complexity, we introduce an innovative evaluators: the LLM-based MD-Judge for QA pairs with a particular focus on attack-enhanced queries, ensuring a seamless, and reliable evaluation. Above components extend SALAD-Bench from standard LLM safety evaluation to both LLM attack and defense methods evaluation, ensuring the joint-purpose utility. Our extensive experiments shed light on the resilience of LLMs against emerging threats and the efficacy of contemporary defense tactics. Data and evaluator are released under https://github.com/OpenSafetyLab/SALAD-BENCH. Warning: this paper includes examples that may be offensive or harmful.

Training Large Language Models (LLMs) with Reinforcement Learning from AI Feedback (RLAIF) aligns model outputs more closely with human preferences. This involves an evaluator model ranking multiple candidate responses to user prompts. However, the rankings from popular evaluator models such as GPT-4 can be inconsistent. We propose the Repeat Ranking method - where we evaluate the same responses multiple times and train only on those responses which are consistently ranked. Using 2,714 prompts in 62 languages, we generated responses from 7 top multilingual LLMs and had GPT-4 rank them five times each. Evaluating on MT-Bench chat benchmarks in six languages, our method outperformed the standard practice of training on all available prompts. Our work highlights the quality versus quantity trade-off in RLAIF dataset generation and offers a stackable strategy for enhancing dataset and thus model quality.

Graphical User Interface (GUI) agents hold great potential for automating complex tasks across diverse digital environments, from web applications to desktop software. However, the development of such agents is hindered by the lack of high-quality, multi-step trajectory data required for effective training. Existing approaches rely on expensive and labor-intensive human annotation, making them unsustainable at scale. To address this challenge, we propose AgentTrek, a scalable data synthesis pipeline that generates high-quality GUI agent trajectories by leveraging web tutorials. Our method automatically gathers tutorial-like texts from the internet, transforms them into task goals with step-by-step instructions, and employs a visual-language model agent to simulate their execution in a real digital environment. A VLM-based evaluator ensures the correctness of the generated trajectories. We demonstrate that training GUI agents with these synthesized trajectories significantly improves their grounding and planning performance over the current models. Moreover, our approach is more cost-efficient compared to traditional human annotation methods. This work underscores the potential of guided replay with web tutorials as a viable strategy for large-scale GUI agent training, paving the way for more capable and autonomous digital agents.

Text-to-image (T2I) models have rapidly advanced, enabling the generation of high-quality images from text prompts across various domains. However, these models present notable safety concerns, including the risk of generating harmful, biased, or private content. Current research on assessing T2I safety remains in its early stages. While some efforts have been made to evaluate models on specific safety dimensions, many critical risks remain unexplored. To address this gap, we introduce T2ISafety, a safety benchmark that evaluates T2I models across three key domains: toxicity, fairness, and bias. We build a detailed hierarchy of 12 tasks and 44 categories based on these three domains, and meticulously collect 70K corresponding prompts. Based on this taxonomy and prompt set, we build a large-scale T2I dataset with 68K manually annotated images and train an evaluator capable of detecting critical risks that previous work has failed to identify, including risks that even ultra-large proprietary models like GPTs cannot correctly detect. We evaluate 12 prominent diffusion models on T2ISafety and reveal several concerns including persistent issues with racial fairness, a tendency to generate toxic content, and significant variation in privacy protection across the models, even with defense methods like concept erasing. Data and evaluator are released under https://github.com/adwardlee/t2i_safety.

Manual curation of evaluation datasets is struggling to keep up with the rapidly expanding capabilities and deployment scenarios of language models. Towards scalable model profiling, we introduce and validate a framework for evaluating LLMs, called Adaptive Evaluations. Adaptive evaluations use scaffolded language models (evaluator agents) to search through a target model's behavior on a domain dataset and create difficult questions (tasks) that can discover and probe the model's failure modes. We find that frontier models lack consistency when adaptively probed with our framework on a diverse suite of datasets and tasks, including but not limited to legal reasoning, forecasting, and online harassment. Generated questions pass human validity checks and often transfer to other models with different capability profiles, demonstrating that adaptive evaluations can also be used to create difficult domain-specific datasets.

Current Vision Language Models (VLMs) remain vulnerable to malicious prompts that induce harmful outputs. Existing safety benchmarks for VLMs primarily rely on automated evaluation methods, but these methods struggle to detect implicit harmful content or produce inaccurate evaluations. Therefore, we found that existing benchmarks have low levels of harmfulness, ambiguous data, and limited diversity in image-text pair combinations. To address these issues, we propose the ELITE benchmark, a high-quality safety evaluation benchmark for VLMs, underpinned by our enhanced evaluation method, the ELITE evaluator. The ELITE evaluator explicitly incorporates a toxicity score to accurately assess harmfulness in multimodal contexts, where VLMs often provide specific, convincing, but unharmful descriptions of images. We filter out ambiguous and low-quality image-text pairs from existing benchmarks using the ELITE evaluator and generate diverse combinations of safe and unsafe image-text pairs. Our experiments demonstrate that the ELITE evaluator achieves superior alignment with human evaluations compared to prior automated methods, and the ELITE benchmark offers enhanced benchmark quality and diversity. By introducing ELITE, we pave the way for safer, more robust VLMs, contributing essential tools for evaluating and mitigating safety risks in real-world applications.

LLM-as-a-Judge has emerged as a promising tool for automatically evaluating generated outputs, but its reliability is often undermined by potential biases in judgment. Existing efforts to mitigate these biases face key limitations: in-context learning-based methods fail to address rooted biases due to the evaluator's limited capacity for self-reflection, whereas fine-tuning is not applicable to all evaluator types, especially closed-source models. To address this challenge, we introduce the Reasoning-based Bias Detector (RBD), which is a plug-in module that identifies biased evaluations and generates structured reasoning to guide evaluator self-correction. Rather than modifying the evaluator itself, RBD operates externally and engages in an iterative process of bias detection and feedback-driven revision. To support its development, we design a complete pipeline consisting of biased dataset construction, supervision collection, distilled reasoning-based fine-tuning of RBD, and integration with LLM evaluators. We fine-tune four sizes of RBD models, ranging from 1.5B to 14B, and observe consistent performance improvements across all scales. Experimental results on 4 bias types--verbosity, position, bandwagon, and sentiment--evaluated using 8 LLM evaluators demonstrate RBD's strong effectiveness. For example, the RBD-8B model improves evaluation accuracy by an average of 18.5% and consistency by 10.9%, and surpasses prompting-based baselines and fine-tuned judges by 12.8% and 17.2%, respectively. These results highlight RBD's effectiveness and scalability. Additional experiments further demonstrate its strong generalization across biases and domains, as well as its efficiency.

World models -- generative models that simulate environment dynamics conditioned on past observations and actions -- are gaining prominence in planning, simulation, and embodied AI. However, evaluating their rollouts remains a fundamental challenge, requiring fine-grained, temporally grounded assessment of action alignment and semantic consistency -- capabilities not captured by existing metrics. Vision-Language Models (VLMs) have shown promise as automatic evaluators of generative content due to their strong multimodal reasoning abilities. Yet, their use in fine-grained, temporally sensitive evaluation tasks remains limited and requires targeted adaptation. We introduce a evaluation protocol targeting two recognition tasks -- action recognition and character recognition -- each assessed across binary, multiple-choice, and open-ended formats. To support this, we present UNIVERSE (UNIfied Vision-language Evaluator for Rollouts in Simulated Environments), a method for adapting VLMs to rollout evaluation under data and compute constraints. We conduct a large-scale study comparing full, partial, and parameter-efficient finetuning across task formats, context lengths, sampling strategies, and data compositions. The resulting unified evaluator matches the performance of task-specific baselines using a single checkpoint. Human studies confirm strong alignment with human judgments, establishing UNIVERSE as a scalable, semantics-aware evaluator for world models.

As Decentralized Finance (DeFi) develops, understanding user intent behind DeFi transactions is crucial yet challenging due to complex smart contract interactions, multifaceted on-/off-chain factors, and opaque hex logs. Existing methods lack deep semantic insight. To address this, we propose the Transaction Intent Mining (TIM) framework. TIM leverages a DeFi intent taxonomy built on grounded theory and a multi-agent Large Language Model (LLM) system to robustly infer user intents. A Meta-Level Planner dynamically coordinates domain experts to decompose multiple perspective-specific intent analyses into solvable subtasks. Question Solvers handle the tasks with multi-modal on/off-chain data. While a Cognitive Evaluator mitigates LLM hallucinations and ensures verifiability. Experiments show that TIM significantly outperforms machine learning models, single LLMs, and single Agent baselines. We also analyze core challenges in intent inference. This work helps provide a more reliable understanding of user motivations in DeFi, offering context-aware explanations for complex blockchain activity.