TCMQA v1 - Multi Answer: leaderboard
TCMQA v1 - Multi Answer. TCMQA arXiv2609.33014v1 Table4,38,279 questions with35,893 single-answer and2,386 multi-answer items; exact-match option-set accuracy. Subset is5,250 practitioner-panel questions. No mean-human score is substituted for the population median. All named baselines retain their instruction/chat generation; no base/instruct merge. Log-probability choice evaluation and multi-answer set decoding as paper Sec2/3. Metric: Exact-set accuracy (%). Higher is better.
Metric: Exact-set accuracy (%). Source: arxiv.org. Saturation forecast: Around February 2027. 27 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Qwen 2.5 14B Instruct | 54.99 | #1732 |
| 2 | Qwen 2 7B Instruct | 50.29 | #3163 |
| 3 | Qwen 3.5 9B | 44.3 | #803 |
| 4 | Qwen 2.5 7B Instruct | 41.66 | #2564 |
| 5 | Qwen 1.5 14B Chat | 36.5 | #2974 |
| 6 | Qwen 3.5 4B | 34.03 | #1175 |
| 7 | Qwen 1.5 7B Chat | 22.88 | #6361 |
| 8 | Qwen 2.5 1.5B Instruct | 22.55 | #7470 |
| 9 | Qwen 2.5 3B Instruct | 21.54 | #5150 |
| 10 | Gemma 3 12B (IT) | 18.11 | #1837 |
| 11 | Qwen 2 1.5B Instruct | 17.23 | #7855 |
| 12 | gemma-4-E4B-it | 15.09 | #1469 |
| 13 | Qwen 2 0.5B Instruct | 14.96 | #9171 |
| 14 | Qwen 3.5 0.8B | 14.92 | #8081 |
| 15 | Gemma 4 E2B (IT) | 14.54 | #2303 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=tcmqa-v1-multi-answer · How It Works · Data refreshed daily, snapshot 2026-11-13.