Chocolatine-14B-Instruct-DPO-v1.2: benchmark results
jpacifico's French-focused DPO fine-tune of Microsoft's Phi-3-medium-4k-instruct, trained on a revised French Orca DPO pairs dataset. Provider: jpacifico. Released 2024-08-12. Access: Open.
Aggregate Rating 1521 ± 44, rank #2058 of 10916 rated models, from 15 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Open Korean LLM - Ko-EQ-Bench | 43.6 | EQ-Bench score | 93.1 |
| Open LLM Leaderboard - BBH | 64.38 | Score | 89.4 |
| Open Korean LLM - Ko-IFEval | 44.96 | Strict accuracy, prompt/instruction mean (%) | 89.3 |
| Open Korean LLM - KorNAT-CKA | 31.86 | Normalized accuracy (%) | 87.8 |
| Open LLM Leaderboard - MMLU-Pro | 46.97 | Score | 85.8 |
| Open LLM Leaderboard - IFEval | 68.52 | Score | 81.2 |
| Open Korean LLM - KorNAT-Harmless | 65.14 | Normalized accuracy (%) | 79.2 |
| Open LLM Leaderboard - GPQA | 32.55 | Score | 78.5 |
| Open LLM Leaderboard - MATH Level 5 | 20.92 | Score | 72.4 |
| Open Korean LLM Leaderboard | 39.1 | Average Score (%) | 69.7 |
| Open LLM Leaderboard - MuSR | 42.68 | Score | 67.3 |
| Open Korean LLM - Ko-GPQA-Diamond | 23.23 | Normalized accuracy (%) | 48.1 |
Interactive version: theaggregate.ai/model?slug=chocolatine-14b-instruct-dpo-v1-2 · How It Works · Data refreshed daily, snapshot 2026-11-08.