MIMeBench: leaderboard
Writing a main-idea multiple-choice item (one correct option and three distractors) for 100 passages from Chinese civil-service aptitude exams; MIMeBench (Li et al., 2026). An LLM judge scores each option out of 10 against item-specific criteria, and the mean item score is out of 40; higher is better.
Metric: Mean item score (0-40; the generated correct main-idea option and three distractors each scored 0-10 by an LLM judge against item-specific criteria, 100 Chinese civil-service exam passages). Source: arxiv.org. Saturation forecast: Estimated already saturated. 3 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Qwen 2.5 7B Instruct | 22.2 | #2583 |
| 2 | DeepSeek-R1-Distill-Qwen-7B | 21.53 | #8045 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=mimebench · How It Works · Data refreshed daily, snapshot 2026-11-12.