ClinTraceBench - Treatment Response: leaderboard
ClinTraceBench (Wang, Zhao and Zhao, Findings of EMNLP 2026) tests longitudinal clinical reasoning over multi-visit patient-clinician dialogues derived from MIMIC-IV records; this task asks whether a lab value rose, fell or stayed similar after a documented treatment, without adjustment for concurrent medication. Accuracy in percent over 134 questions with the whole dialogue history in context; higher is better.
Metric: Accuracy (%; task T8, observed treatment response: whether a lab rose, fell or stayed similar after a documented treatment, 134 questions; full-context strategy: the whole multi-visit patient-clinician dialogue derived from MIMIC-IV records is given verbatim, models called through OpenRouter). Source: arxiv.org. Saturation forecast: Around February 2027. 4 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Sonnet 4.6 | 47 | #158 |
| 2 | Claude Haiku 4.5 | 46.3 | #568 |
| 3 | DeepSeek V3.1 | 44 | #500 |
| 4 | GPT-4o Mini | 26.9 | #1627 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=clintracebench-treatment-response · How It Works · Data refreshed daily, snapshot 2026-11-12.