AgentProcessBench: Diagnosing Step-Level Process Quality in Tool-Using Agents Paper • 2603.14465 • Published 3 days ago • 14
HyPER: Bridging Exploration and Exploitation for Scalable LLM Reasoning with Hypothesis Path Expansion and Reduction Paper • 2602.06527 • Published Feb 6
DARC: Decoupled Asymmetric Reasoning Curriculum for LLM Evolution Paper • 2601.13761 • Published Jan 20 • 16