MedConclusion - Non-Contradiction: leaderboard
Metric: Non-contradiction rate (judge score 0-100): the model writes a formal academic conclusion (prompt A, no length or style constraint) from the non-conclusion sections of a PubMed structured abstract; a GPT-5.4-mini judge scores it 0-100 against the author-written conclusion, averaged over a random 30K-abstract subset of MedConclusion; temperature 0, at most 1,024 new tokens; higher is better. Source: arxiv.org. Saturation forecast: Around January 2027. 15 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-5.4 (Non-reasoning) | 84.61 |
| 2 | Gemini 3.1 Pro (Preview) | 82.02 |
| 3 | MiniMax-M2.1 | 81.89 |
| 4 | Gemma 3 27B (IT) | 81.55 |
| 5 | Kimi K2 (Thinking) | 80.92 |
| 6 | GLM-4.6V | 80.5 |
| 7 | DeepSeek V3.2 (Non-reasoning) | 80.31 |
| 8 | Llama 3.1 8B Instruct | 80.24 |
| 9 | DeepSeek R1 | 79.67 |
| 10 | Gemma 2 9B (IT) | 79.12 |
| 11 | Qwen 3 4B 2507 Instruct | 78.96 |
| 12 | Qwen 2.5 VL 7B Instruct | 78.73 |
| 13 | Qwen 2.5 7B Instruct | 77.5 |
| 14 | Llama 3.2 1B Instruct | 66.14 |
Interactive version: theaggregate.ai/benchmark?slug=medconclusion-non-contradiction · How It Works · Data refreshed daily, snapshot 2026-10-07.