AI search agents fail due to query ambiguity, not search capabilities
A benchmark (DiscoBench) shows iterative-searching AI models without clarification achieve only 51.9% accuracy, while the best models average 43%.
Published 1sem1 sourceNotable
Lire en français
≈ 22s
43 %
9% accuracy, while the best models average 43%
The fact
Query ambiguity severely impacts performance: accuracy jumps by 40 points when removed.
AI agents struggle more with asking follow-up questions than with the search process itself.
Click the link to read an article on the topic: