PCoA - Summary: leaderboard
PCoA (Chu et al., University of Duisburg-Essen, 2026) asks a model to summarize one of sixteen medical aspects (such as participants, intervention, outcomes, blinding or funding) of a randomized controlled trial abstract, cite the supporting sentences and extract the contributory phrases from them, over 1,799 expert-annotated instances, zero-shot. This board scores the summary: claim F1 (0 to 1) of claim recall against the reference summary's atomic claims and claim precision of the generated claims, with claims split by Mistral-Large-2411 and entailment checked by TRUE; higher is better.
Metric: Claim F1 (0-1; harmonic mean of claim recall against reference subclaims and claim precision of generated subclaims). Source: arxiv.org. Saturation forecast: Estimated already saturated. 4 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | DeepSeek V3 (0324) | 0.66 | #704 |
| 2 | Mistral Large 2 (Nov) Instruct (2411) | 0.65 | #1199 |
| 3 | GPT-4o | 0.62 | #741 |
| 4 | Llama 3.1 70B Instruct | 0.61 | #1541 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=pcoa-summary · How It Works · Data refreshed daily, snapshot 2026-11-12.