Benchmark
ScienceQA
Science benchmarks measure how well models answer science questions. Scores are percentages.
All science benchmarksScores from Epoch AI Benchmarking Hub data, LLM Stats data.
- Models measured
- 7
- Scores recorded
- 7
- Sources
- 2
- IQ weight
- 0.76
Score distribution
- 69.9%
- Mean
- 74%
- Median
- 19.1%
- Std dev
- 43.3%
- 10th percentile
- 91.3%
- 90th percentile
- 43.1%
- Min
This benchmark has a weight of 0.76 for chat models, measured over 7 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 |
|---|---|---|
| 1 | Phi-3.5 Vision Instruct· Epoch AI Benchmarking Hub dataValues from 2 sources
| 91.3% |
| 2 | GPT 4o· Epoch AI Benchmarking Hub data | 88.5% |
| 3 | text-davinci-001· Epoch AI Benchmarking Hub data | 74% |
| 4 | Claude 3 Haiku· Epoch AI Benchmarking Hub data | 72% |
| 5 | Llama 2 13b· Epoch AI Benchmarking Hub data | 55.8% |
| 6 | LLaMA 13B· Epoch AI Benchmarking Hub data | 43.3% |
| 7 | Llama 2 7b· Epoch AI Benchmarking Hub data | 43.1% |
Data credits: Epoch AI Benchmarking Hub, CC BY 4.0 (licence) · Data by LLM Stats