Study Finds Search Agents Rely on Tools, Drop Sharply on Non-English Queries
A new paper questions the validity of agentic search benchmarks, finding Sol 5.6 answers 46% of test questions without any search tools. Users also found search agent performance drops sharply when queries are in non-English languages.
2026-08-28 ~ 2026-08-28 · 2 related posts
- Paper: Sol 5.6 Achieves 46% Accuracy on BrowseComp-Plus Without Search — beirmug · 2026-08-28
- Research finds search agents drop performance on non-English queries — beirmug · 2026-08-28