Study Finds LLMs Underestimate Missing Information; Question Order Affects Diagnostic Accuracy
A joint study by Harvard Medical School, Broad Institute, Kempner Institute and Google DeepMind formalizes multi-turn information seeking as k-underdetermined constraint satisfaction and introduces the MT-INFOSEEK benchmark. It finds that LLMs can detect but severely underestimate missing information, and that question order affects diagnostic accuracy.
2026-08-19 ~ 2026-08-19 · 3 related posts
- Study: Can LLMs recognize missing info and ask questions? — marinkazitnik · 2026-08-19
- Paper: LLMs severely underestimate missing information, study finds — marinkazitnik · 2026-08-19
- Study: question order affects AI diagnostic accuracy; final accuracy doesn't measure info seeking — marinkazitnik · 2026-08-19