Clinical Verbatim Citation - Claim Support (Strict): leaderboard
Of the clinical-QA claims with a verbatim-compliant quote, the share whose quote fully supports the claim as judged by deepseek-v4-flash. Higher is better. Source: arxiv.org/abs/2609.15964.
Metric: Fully supported cited claims (%, judge deepseek-v4-flash). Source: arxiv.org. Saturation forecast: Not forecast. 12 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5 Mini | 79 | #193 |
| 2 | GPT-5.4 | 76.9 | #83 |
| 3 | DeepSeek V4 Pro | 73.8 | #112 |
| 4 | GPT-5.4 Mini | 62.4 | #228 |
| 5 | DeepSeek V4 Flash | 60.5 | #155 |
| 6 | Claude Haiku 4.5 | 44.1 | #287 |
| 7 | GLM-5.2 | 40.3 | #80 |
| 8 | Claude Sonnet 5 | 38.5 | #65 |
| 9 | GLM-4.7 Flash | 38.2 | #543 |
| 10 | Claude Opus 5 | 37.8 | #16 |
| 11 | GPT-4.1 Mini | 36.2 | #363 |
| 12 | Kimi K3 | 33.8 | #35 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=clinical-verbatim-citation-claim-support-strict · How It Works · Data refreshed daily, snapshot 2026-11-05.