CMNIE - Argument Classification: leaderboard
Zero-shot event extraction from Chinese military news: argument classification, exact-match F1. Higher is better. Source: arxiv.org/abs/2609.10722.
Metric: F1 (%). Source: arxiv.org. Saturation forecast: Around 2030. 11 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.1 | 31.8 | #145 |
| 2 | Qwen 2.5 32B Instruct | 29.7 | #509 |
| 3 | Llama 3.1 70B Instruct | 29.5 | #567 |
| 4 | Llama 3.3 70B Instruct | 29.4 | #541 |
| 5 | Qwen 2.5 72B Instruct | 29.3 | #469 |
| 6 | Moonshot v1 8K | 27.8 | #622 |
| 7 | Qwen 2.5 14B Instruct | 26.7 | #654 |
| 8 | Qwen 2.5 7B Instruct | 26.3 | #863 |
| 9 | Llama 3.1 8B Instruct | 23.1 | #1017 |
| 10 | GLM-4 9B Chat | 20.8 | #911 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=cmnie-argument-classification · How It Works · Data refreshed daily, snapshot 2026-11-05.