MiGUE-Bench - Event Detection: leaderboard
The event detection task of MiGUE-Bench (2026), a benchmark of event analysis over news documents: micro-F1 in percent on 330 event-trigger questions posed as multiple choice with several correct options. Higher is better.
Metric: Micro-F1 (%) on 330 event-trigger detection questions (multiple-choice questions with LLM-built distractors over news documents, several correct options); higher is better. Source: arxiv.org. Saturation forecast: Around February 2027. 12 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Opus 4.5 | 85.52 | #156 |
| 2 | GPT-5.2 Pro | 83.98 | #124 |
| 3 | DeepSeek V3.2 | 81.68 | #406 |
| 4 | GLM-4.7 | 81.33 | #366 |
| 5 | Gemini 3 Pro | 71.51 | #137 |
| 6 | Qwen 3 Max | 65.43 | #403 |
| 7 | Qwen 3 30B A3B | 62.98 | #1122 |
| 8 | Qwen 3 8B | 61.38 | #1830 |
| 9 | Qwen 3 235B A22B | 61.15 | #623 |
| 10 | Claude Haiku 4.5 | 61.01 | #568 |
| 11 | Kimi K2 | 58.94 | #439 |
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-event-detection · How It Works · Data refreshed daily, snapshot 2026-11-13.