KNOWS Benchmark: Top Web Agents Under 3% on Complex Long Tasks

Researchers including a University of Utah team released KNOWS, a benchmark evaluating web agents on open-ended, multi-step workflows that require synthesizing retrieved information into final deliverables. The best agents achieve less than 3% full success on these complex long tasks.

2026-10-18 ~ 2026-10-18 · 2 related posts

• KNOWS Benchmark: Best Web Agent Fully Succeeds on Fewer Than 3% of Complex Tasks — anmarasovic · 2026-10-18
• KNOWS benchmark: best web agent fully succeeds in fewer than 3% of complex long-horizon tasks — anmarasovic · 2026-10-18