OSWorld

The 'OSWorld' dataset/benchmark contains a variety of tasks used to evaluate the performance of reinforcement learning systems, specifically in the context of large language models and agentic scenarios.

Papers using OSWorld (10)

Reinforcement Learning for Computer-Use Agents with Autonomous Evaluation2026IntentScore: Intent-Conditioned Action Evaluation for Computer-Use Agents2026RLAnything: Forge Environment, Policy, and Reward Model in Completely Dynamic RL System2026OS-Oracle: A Comprehensive Framework for Cross-Platform GUI Critic Models2025STEP: Success-Rate-Aware Trajectory-Efficient Policy Optimization2025UI-TARS-2 Technical Report: Advancing GUI Agent with Multi-Turn Reinforcement Learning2025Efficient Multi-turn RL for GUI Agents via Decoupled Training and Adaptive Data Curation2025ComputerRL: Scaling End-to-End Online Reinforcement Learning for Computer Use Agents2025Mobile-Agent-v3: Fundamental Agents for GUI Automation2025ZeroGUI: Automating Online GUI Learning at Zero Human Cost2025