HELM Lite - LegalBench - Function of Decision Section: leaderboard
HELM Lite v1.13.0 LegalBench Function of Decision Section task: label a paragraph of a judicial opinion with its function (facts, procedural history, issue, rule, analysis, conclusion or decree); 367 items per model, exact match. Percent correct, 0-100, higher is better.
Metric: Exact Match (%). Source: crfm.stanford.edu. Saturation forecast: Estimated already saturated. 91 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Llama 3 70B | 46.59 | #632 |
| 2 | Qwen 2.5 72B Instruct | 46.05 | #469 |
| 3 | Gemini 1.5 Pro (001) | 46.05 | #465 |
| 4 | Llama 3.1 70B Instruct | 45.78 | #567 |
| 5 | Claude 3.5 Sonnet (20240620) | 45.5 | #308 |
| 6 | Llama 3.2 90B Vision Instruct | 45.23 | #621 |
| 7 | GPT-4 (0613) | 45.23 | #493 |
| 8 | Llama 3.1 405B Instruct | 44.96 | #470 |
| 9 | Llama 2 70B | 44.41 | #1039 |
| 10 | Nova Pro | 44.41 | #594 |
| 11 | GPT-4o (2024-05-13) | 44.14 | #337 |
| 12 | Mixtral 8x22B | 44.14 | #725 |
| 13 | Gemini 1.5 Pro (002) | 43.87 | #368 |
| 14 | Gemma 2 27B (IT) | 43.87 | #652 |
| 15 | Mistral-7B-v0.1 | 43.32 | #1294 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=helm-lite-legalbench-function-of-decision-section · How It Works · Data refreshed daily, snapshot 2026-11-05.