gemma-2-9B-it-DPO: benchmark results
Princeton NLP's DPO fine-tune of Gemma 2 9B IT on ultrafeedback-armorm, the DPO baseline counterpart to their SimPO model. Provider: Princeton NLP. Released 2024-07-16. Access: Open.
Aggregate Rating 1531 ± 19, rank #678 of 1656 rated models, from 24 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Open Korean LLM - Ko-GSM8K (flexible extract) | 67.25 | Exact match, flexible extract (%) | 96.8 |
| Open Korean LLM - Ko-IFEval | 54.42 | Strict accuracy, prompt/instruction mean (%) | 96.2 |
| AlpacaEval 2.0 | 67.66 | LC Win Rate (%) | 95 |
| Open Korean LLM - KorNAT-CKA | 33.8 | Normalized accuracy (%) | 93.3 |
| Open Portuguese LLM - ASSIN2 RTE | 93.45 | Macro F1 (%) | 89.5 |
| WildBench | 53.22 | WB Score Task-Macro | 85.5 |
| BenchBench | 81.01 | Aggregate Score (%) | 85.3 |
| Open LLM Leaderboard - GPQA | 33.56 | Score | 82.7 |
| Open LLM Leaderboard - BBH | 59.41 | Score | 80.9 |
| Open Korean LLM - Ko-Winogrande | 63.14 | Accuracy (%) | 72.6 |
| Open Portuguese LLM - BLUEX | 57.16 | Accuracy (%) | 68.6 |
| Open LLM Leaderboard - MMLU-Pro | 37.23 | Score | 61.2 |
Interactive version: theaggregate.ai/model?slug=gemma-2-9b-it-dpo · How It Works · Data refreshed daily, snapshot 2026-10-20.