HELM SeaHELM - LINDSEA Presuppositions (id): leaderboard
Metric: EM. Source: crfm.stanford.edu. 21 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Llama 3.3 70B Instruct | 80 |
| 2 | Qwen 2.5 72B Instruct | 80 |
| 3 | Llama 3.1 70B Instruct | 78.75 |
| 4 | Llama 3.2 90B Vision Instruct | 73.75 |
| 5 | Qwen 2 72B Instruct | 72.5 |
| 6 | Claude 3.7 Sonnet (20250219) | 71.25 |
| 7 | Llama 3.1 405B Instruct | 70 |
| 8 | DeepSeek V3 | 67.5 |
| 9 | Gemini 2.0 Flash | 66.25 |
| 10 | Gemini 2.0 Flash Lite | 66.25 |
| 11 | GPT-4o (2024-11-20) | 62.5 |
| 12 | Llama 4 Maverick Instruct FP8 | 60 |
| 13 | Llama 4 Scout Instruct | 56.25 |
| 14 | DeepSeek R1 Distill Llama 8B | 50 |
| 15 | GPT-4o Mini (2024-07-18) | 48.75 |
Interactive version: theaggregate.ai/benchmark?slug=helm-seahelm-lindsea-presuppositions-id · How It Works · Data refreshed daily, snapshot 2026-09-05.