Structuring MoE Expert Selection for Agentic Reinforcement Learning Paper • 2610.07332 • Published 20 days ago • 4
Stepped MoE: Segment-Level Routing with Configurable Inference Complexity Paper • 2610.07348 • Published 20 days ago • 5
Decoding Looped Transformers Better for (Almost) Free Paper • 2610.02185 • Published 24 days ago • 48
It Takes Two to Match: Co-Evolving Generative Retriever with Reinforcement Learning Paper • 2609.00638 • Published Sep 1 • 67
Beyond Visual CoT: Internalized Visual Thinking for Proactive Video Reasoning Paper • 2608.15869 • Published Aug 16 • 11
Environment-free Synthetic Data Generation for API-Calling Agents Paper • 2607.16900 • Published Jul 18 • 21
TopoPrimer: The Missing Topological Context in Forecasting Models Paper • 2605.15035 • Published May 14 • 3
Unmasking On-Policy Distillation: Where It Helps, Where It Hurts, and Why Paper • 2605.10889 • Published May 11 • 6
A Single Neuron Is Sufficient to Bypass Safety Alignment in Large Language Models Paper • 2605.08513 • Published May 8 • 18