EVALITA - relation-extraction: leaderboard
The "relation-extraction" subtask of EVALITA. A task from EVALITA, the periodic evaluation campaign for Italian-language natural language processing and understanding.
Metric: CPS. Source: huggingface.co. Saturation forecast: Around April 2028. 48 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Mistral Large 2 (Nov) Instruct (2411) | 66.82 | #465 |
| 2 | Llama-3.1-Tulu-3-70B | 55.38 | #482 |
| 3 | Gemma 2 9B (IT) | 51.26 | #788 |
| 4 | Gemma 3n E4B (IT) | 49.65 | #749 |
| 5 | Gemma 3 27B (IT) | 48.48 | #523 |
| 6 | Qwen2.5-14B-Instruct-1M | 44.23 | #578 |
| 7 | LLaMAntino-3-ANITA-8B-Inst-DPO-ITA | 42.14 | #1009 |
| 8 | Qwen 2.5 72B Instruct | 41.6 | #454 |
| 9 | Gemma 3 12B (IT) | 41.59 | #664 |
| 10 | Mistral Small 3 | 40.9 | #627 |
| 11 | granite-3.1-8B-instruct | 39.91 | #1069 |
| 12 | Qwen 3 Next 80B A3B Instruct | 38.87 | #368 |
| 13 | Qwen 2.5 VL 32B Instruct | 38.41 | #453 |
| 14 | DeepSeek R1 Distill Llama 70B | 37.34 | #594 |
| 15 | Phi-4 | 37.15 | #706 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=evalita-relation-extraction · How It Works · Data refreshed daily, snapshot 2026-10-20.