Kolab dataset benchmarks human-agent collaboration in professional workflows
October 14, 2026
The Kolab dataset provides 115+ hours of screen recordings documenting 50 professionals collaborating with AI agents on long-horizon tasks. It shifts benchmarking focus from pure automation to effective human-agent interaction and plan revision.
HOW THIS AFFECTS YOU
●
builderYou can use this data to optimize agentic workflows for real-world human-in-the-loop scenarios.
●
researcherThis provides a more realistic benchmark for evaluating how agents perform in interdependent professional environments.