Faro-Yi-9B-DPO: benchmark results
Wenbo Pan's DPO-tuned Faro-Yi-9B, a long-context Yi-9B derivative preference-trained on UltraFeedback, Orca, and Truthy DPO pairs. Provider: wenbopan. Released 2024-04-07. Access: Open.
Aggregate Rating 1427 ± 15, rank #1115 of 1656 rated models, from 41 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Open LLM Leaderboard v1 - MMLU | 70.74 | Accuracy (%) (5-shot) | 94.3 |
| Open CoT - LSAT Logical Reasoning | 18.24 | CoT Gain (%) | 85.9 |
| Open Korean LLM - Ko-EQ-Bench | 34.35 | EQ-Bench score | 84.2 |
| Open CoT - LSAT Reading Comprehension | 19.7 | CoT Gain (%) | 84 |
| Open CoT - LogiQA 2 | 12.91 | CoT Gain (%) | 83.2 |
| Open Chinese LLM Leaderboard | 64.43 | Official Nonzero-Task Mean (%) | 80.2 |
| Open Chinese LLM - WinoGrande | 66.69 | Accuracy (%) | 80 |
| Open Chinese LLM - C-Eval Semantic | 86.22 | Accuracy (%) | 79.3 |
| Open LLM Leaderboard v1 - GSM8K | 64.75 | Accuracy (%) (5-shot) | 78.9 |
| Open Chinese LLM - ARC Challenge | 57.76 | Accuracy (%) | 76.6 |
| Open Chinese LLM - HellaSwag | 64.24 | Accuracy (%) | 76 |
| Open Korean LLM - KorNAT-Harmless | 64.64 | Normalized accuracy (%) | 76 |
Interactive version: theaggregate.ai/model?slug=faro-yi-9b-dpo · How It Works · Data refreshed daily, snapshot 2026-10-20.