Benchmark
HELM Lite: MATH - Equivalent (CoT)
Math benchmarks measure how well models solve maths problems. Scores are percentages. Higher is better.
All math benchmarksScores from Stanford HELM Lite data.
- Models measured
- 69
- Scores recorded
- 70
- Sources
- 1
- IQ weight
- 1.57
Score distribution
- 59.2%
- Mean
- 67.4%
- Median
- 24.7%
- Std dev
- 23.6%
- 10th percentile
- 83.5%
- 90th percentile
- 2.9%
- Min
This benchmark has a weight of 1.57 for chat models, measured over 69 models. The weight is how much it counts toward llminfo IQ. We measure it from the data, not from the raw scores.
Leaderboard
| # | Model | Score |
|---|---|---|
| 51 | Text Davinci 002· Stanford HELM Lite data | 42.8% |
| 52 | Command R Plus· Stanford HELM Lite data | 40.3% |
| 53 | Meta Llama 3 8b· Stanford HELM Lite data | 39.1% |
| 54 | Jamba Instruct· Stanford HELM Lite data | 38% |
| 55 | Yi-34B· Stanford HELM Lite data | 37.5% |
| 56 | DBRX-instruct· Stanford HELM Lite data | 35.8% |
| 57 | Llama 2 70b· Stanford HELM Lite data | 32.3% |
| 58 | Jamba 1.5 Mini· Stanford HELM Lite data | 31.8% |
| 59 | Mistral 7B v0.1· Stanford HELM Lite data | 29.7% |
| 60 | Mistral (7B) Instruct v0.3· Stanford HELM Lite data | 28.9% |
| 61 | Command R+· Stanford HELM Lite data | 26.6% |
| 62 | phi-2· Stanford HELM Lite data | 25.5% |
| 63 | command· Stanford HELM Lite data | 23.6% |
| 64 | Claude 3 Haiku· Stanford HELM Lite data | 13.1% |
| 65 | Yi-6B· Stanford HELM Lite data | 12.6% |
| 66 | Llama 2 13b· Stanford HELM Lite data | 10.2% |
| 67 | Command Light· Stanford HELM Lite data | 9.8% |
| 68 | Llama 2 7b· Stanford HELM Lite data | 9.7% |
| 69 | Claude 3 Sonnet· Stanford HELM Lite data | 8.4% |
| 70 | OLMo-7B· Stanford HELM Lite data | 2.9% |
Showing 51–70 of 70 results.
Data credits: Stanford HELM (Stanford CRFM), Apache-2.0 (licence)