Seonglae ChoDefine step -> Collect data -> Make evals
- End-to-End is not just a sum of components of eval pipeline
- Hard to obey for Good benchmark principles
Computer Use Benchmarks
Browser Use Benchmarks
AI Agent Reasoning Benchmarks
Tool learning Benchmarks
GUI grounding Benchmarks
CodeMidas: Open source as RL env
Agent Benchmarks
Cost-controlled evaluations and joint optimization of accuracy and cost
Stagehand GUI grounding Visual Grounding
OS-Harm benchmark
BrowseSafe: AI browser-specific 'real-time malicious prompt detection model'
