HELM EWoK - Quantitative Properties: leaderboard
HELM EWoK (Elements of World Knowledge, EWoK-CORE-1.0; release v1.0.0) quantitative properties domain: each target sentence is shown with two minimal-pair contexts and the model must pick the context that makes it plausible; 380 items, exact match. Percent correct, 0-100, higher is better; chance is 50.
Metric: Exact Match (%). Source: crfm.stanford.edu. Saturation forecast: Estimated already saturated. 15 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Qwen 2 72B Instruct | 100 | #579 |
| 2 | Claude 3 Opus (20240229) | 100 | #428 |
| 3 | Claude 3.5 Sonnet (20240620) | 99.47 | #308 |
| 4 | GPT-4 Turbo | 98.68 | #417 |
| 5 | Claude 3 Sonnet (20240229) | 98.42 | #694 |
| 6 | GPT-4o (2024-05-13) | 98.16 | #337 |
| 7 | Llama 3 70B Instruct | 98.16 | #640 |
| 8 | Mixtral 8x22B Instruct | 98.16 | #738 |
| 9 | GPT-4o Mini (2024-07-18) | 97.63 | #568 |
| 10 | Claude 3 Haiku (20240307) | 96.84 | #780 |
| 11 | Mixtral 8x7B Instruct | 92.89 | #1085 |
| 12 | Mistral 7B Instruct (v0.3) | 91.32 | #1255 |
| 13 | GPT-3.5 Turbo (0125) | 90.79 | #975 |
| 14 | Claude 3.5 Haiku (20241022) | 90.53 | #590 |
| 15 | Llama 3 8B Instruct | 17.63 | #1116 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=helm-ewok-quantitative-properties · How It Works · Data refreshed daily, snapshot 2026-11-05.