Kolab Benchmark Tests Human-AI Collaboration with 115 Hours of Real Video

The Kolab benchmark, built from 115+ hours of real screen recordings of 50 professionals across 40 occupations, tests human-AI agent collaboration in long-horizon workflows. It finds that frontier multimodal models overestimate the quality of their own collaboration.

2026-10-27 ~ 2026-10-28 · 2 related posts

• Kolab Benchmark Uses Pro Screen Recordings to Show AI Agents Overestimate Collaboration Quality — minnesotanlp · 2026-10-27
• Kolab benchmark: 115+ hours of human-AI collaboration videos across 40 occupations — windx0303 · 2026-10-28