OSReward: Instituting Standardized Evaluation for Cross-Platform Computer-Use Reward Models Paper • 2607.28609 • Published Jul 30 • 69
OSWorld2.0: Benchmarking Computer Use Agents on Long-Horizon Real-World Tasks Paper • 2606.29537 • Published Jun 28 • 24
AgentHorizon: Evaluating Agentic Judges for Long-Horizon Computer-Use Tasks Paper • 2610.11050 • Published Oct 8 • 1
CoTrace: Data Recipes for Training Terminal Agents with Harness-Model Co-Evolution Paper • 2610.10426 • Published Oct 7 • 12
Verifier-Backed Hard Problem Generation for Mathematical Reasoning Paper • 2605.06660 • Published May 7 • 1
Step-Level Sparse Autoencoder for Reasoning Process Interpretation Paper • 2603.03031 • Published Mar 3