Benchmark
HREF Generation
Instruction following benchmarks measure how closely models follow instructions. Scores are percentages.
- Models measured
- 14
- Scores recorded
- 14
- Sources
- 1
- IQ weight
- 0.93
Score distribution
- 29.5%
- Mean
- 32.7%
- Median
- 13.2%
- Std dev
- 11%
- 10th percentile
- 45.4%
- 90th percentile
- 7.2%
- Min
This benchmark has a weight of 0.93 for chat models, measured over 14 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 | Mistral-Large-Instruct-2407· HREF (Ai2) data | 49.2% |
| 2 | Qwen2.5 72B Instruct· HREF (Ai2) data | 45.4% |
| 3 | Llama 3.1 70B Instruct· HREF (Ai2) data | 44.5% |
| 4 | Qwen1.5-110B-Chat· HREF (Ai2) data | 41.4% |
| 5 | Llama 3.1 8B Instruct· HREF (Ai2) data | 39.8% |
| 6 | Yi 1.5 34B Chat· HREF (Ai2) data | 35.3% |
| 7 | Qwen2 72B Instruct· HREF (Ai2) data | 32.7% |
| 8 | Phi-3 Medium 4K Instruct· HREF (Ai2) data | 28.6% |
| 9 | Llama-2-70b-chat-hf· HREF (Ai2) data | 23.9% |
| 10 | Mistral (7B) Instruct v0.3· HREF (Ai2) data | 19% |
| 11 | Llama-2-13b-chat-hf· HREF (Ai2) data | 17.9% |
| 12 | Llama-2-7b-chat-hf· HREF (Ai2) data | 16.6% |
| 13 | vicuna-13b-v1.5· HREF (Ai2) data | 11% |
| 14 | vicuna-7b-v1.5· HREF (Ai2) data | 7.2% |