DALPHIN (Contextual) - Organ Recognition: leaderboard
Naming the organ or residual normal tissue shown in histopathology images (a low-resolution slide overview plus regions of interest) in the DALPHIN pathology copilot benchmark (Lems et al., 2026), on the 272 of its 300 cases where tissue is identifiable. Answers are scored on a hierarchical tissue taxonomy (1 for an exact match, 0.75 or 0.5 for near misses, else 0) on a 0-1 scale; higher is better. Contextual scenario: each case's questions are asked in sequence within one conversation, so earlier answers stay in context.
Metric: Organ recognition score (0-1; free-response organ or residual normal tissue named from low-resolution overviews and regions of interest, scored on a hierarchical tissue taxonomy: 1 exact, 0.75 one step away, 0.5 two steps away, else 0; 272 of 300 cases; contextual scenario: all questions of a case asked in sequence in one conversation). Source: arxiv.org. Saturation forecast: Around February 2027. 3 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 2.5 Pro (Medium) | 0.73 | #262 (Gemini 2.5 Pro) |
| 2 | GPT-5 (Medium) | 0.44 | #163 (GPT-5) |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=dalphin-contextual-organ-recognition · How It Works · Data refreshed daily, snapshot 2026-11-12.