MiGUE-Bench - Intra-Document Coreference: leaderboard
A within-document relation task of MiGUE-Bench (2026), a benchmark of event analysis over news documents: accuracy in percent on 300 two-option event coreference questions about one document. Higher is better.
Metric: Accuracy (%) on 300 single-document event coreference questions (2 options); higher is better. Source: arxiv.org. Saturation forecast: Estimated already saturated. 12 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.2 Pro | 97.35 | #124 |
| 2 | Claude Opus 4.5 | 97 | #156 |
| 3 | GLM-4.7 | 96.33 | #366 |
| 4 | DeepSeek V3.2 | 96.25 | #406 |
| 5 | Gemini 3 Pro | 95 | #137 |
| 6 | Claude Haiku 4.5 | 93 | #568 |
| 7 | Kimi K2 | 92 | #439 |
| 8 | Qwen 3 Max | 90.5 | #403 |
| 9 | Qwen 3 235B A22B | 87.25 | #623 |
| 10 | Qwen 3 8B | 86.77 | #1830 |
| 11 | Qwen 3 30B A3B | 62.11 | #1122 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=migue-bench-intra-document-coreference · How It Works · Data refreshed daily, snapshot 2026-11-13.