HELM Classic - Dyck: leaderboard
The "Dyck" subtask of HELM Classic. A scenario from Stanford HELM Classic, the holistic evaluation suite scoring language models across accuracy, robustness and calibration on a broad set of tasks.
Metric: Exact Match (%). Source: crfm.stanford.edu. Saturation forecast: Estimated already saturated. 69 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | pythia-12B | 79.8 | #1621 |
| 2 | text-davinci-003 | 75.07 | #1146 |
| 3 | gpt-neox-20B | 74.73 | #1600 |
| 4 | Llama 2 70B | 74.2 | #1201 |
| 5 | RedPajama-INCITE-Instruct-3B-v1 | 70.8 | #1617 |
| 6 | LLaMA-30B | 70.6 | #1423 |
| 7 | Mistral-7B-v0.1 | 69.6 | #1326 |
| 8 | vicuna-13B-v1.3 | 68.6 | #1447 |
| 9 | LLaMA-13B | 68.6 | #1516 |
| 10 | davinci | 66.8 | #1553 |
| 11 | Llama 2 7B | 65.8 | #1520 |
| 12 | pythia-6.9B | 65.2 | #1646 |
| 13 | LLaMA-65B | 65 | #1303 |
| 14 | mpt-30B | 63.4 | #1476 |
| 15 | Llama 2 13B | 62.4 | #1436 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=helm-classic-dyck · How It Works · Data refreshed daily, snapshot 2026-10-20.