Benchmark
HELM Lite: NarrativeQA - F1
Scores are percentages. Higher is better.
All general benchmarksScores from Stanford HELM Lite data.
- Models measured
- 69
- Scores recorded
- 70
- Sources
- 1
- IQ weight
- 0.96
Score distribution
- 69.5%
- Mean
- 74.2%
- Median
- 12.7%
- Std dev
- 58.1%
- 10th percentile
- 78.3%
- 90th percentile
- 11.1%
- Min
This benchmark has a weight of 0.96 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 | Yi-6B· Stanford HELM Lite data | 70.2% |
| 52 | Llama 2 7b· Stanford HELM Lite data | 68.6% |
| 53 | Claude 2.1· Stanford HELM Lite data | 67.7% |
| 54 | Jamba 1.5 Large· Stanford HELM Lite data | 66.4% |
| 55 | Jamba Instruct· Stanford HELM Lite data | 65.8% |
| 56 | GPT-3.5 Turbo· Stanford HELM Lite data | 65.5% |
| 57 | Snowflake Arctic Instruct· Stanford HELM Lite data | 65.4% |
| 58 | Command Light· Stanford HELM Lite data | 62.9% |
| 59 | Claude Instant 1.2· Stanford HELM Lite data | 61.6% |
| 60 | Qwen1.5-72B· Stanford HELM Lite data | 60.1% |
| 61 | OLMo-7B· Stanford HELM Lite data | 59.7% |
| 62 | text-unicorn· Stanford HELM Lite data | 58.3% |
| 63 | DeepSeek LLM 67B Chat· Stanford HELM Lite data | 58.1% |
| 64 | Mistral Small· Stanford HELM Lite data | 51.9% |
| 65 | DBRX-instruct· Stanford HELM Lite data | 48.8% |
| 66 | Mistral Large· Stanford HELM Lite data | 45.4% |
| 67 | Qwen1.5-7B· Stanford HELM Lite data | 44.8% |
| 68 | Claude 3 Opus· Stanford HELM Lite data | 35.1% |
| 69 | Claude 3 Haiku· Stanford HELM Lite data | 24.4% |
| 70 | Claude 3 Sonnet· Stanford HELM Lite data | 11.1% |
Showing 51–70 of 70 results.
Data credits: Stanford HELM (Stanford CRFM), Apache-2.0 (licence)