Diagnosing Harmful Continuation in Answer-Correct Long-CoT Training Traces Paper • 2605.29288 • Published 7 days ago • 9
MARS: Enabling Autoregressive Models Multi-Token Generation Paper • 2604.07023 • Published Apr 8 • 38
DeepResearchEval: An Automated Framework for Deep Research Task Construction and Agentic Evaluation Paper • 2601.09688 • Published Jan 14 • 127