Qwen2.5-14B-Gutenberg-1e-Delta: benchmark results
v000000's DPO fine-tune of Qwen2.5 14B Instruct on the Gutenberg literary-prose preference dataset. Provider: v000000. Released 2024-09-20. Access: Open.
Aggregate Rating 1591 ± 17, rank #473 of 1656 rated models, from 144 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Open Ko-LLM Leaderboard | 54.8 | Average (%) | 99.7 |
| Open Korean LLM Leaderboard | 54.8 | Average Score (%) | 99.7 |
| Open Korean LLM - Ko-GSM8K (flexible extract) | 76.19 | Exact match, flexible extract (%) | 99.3 |
| Open Korean LLM - Ko-IFEval | 68.07 | Strict accuracy, prompt/instruction mean (%) | 99.1 |
| Open Korean LLM - KorNAT-CKA | 39.68 | Normalized accuracy (%) | 98.9 |
| Open Korean LLM - Ko-EQ-Bench | 59.25 | EQ-Bench score | 98.6 |
| Open Korean LLM - Ko-GPQA-Diamond | 28.79 | Normalized accuracy (%) | 98.1 |
| Open LLM Leaderboard - MATH Level 5 | 52.64 | Score | 98 |
| Open Korean LLM - KorNAT-Helpful | 60.2 | Normalized accuracy (%) | 97.8 |
| Open LLM Leaderboard - IFEval | 80.45 | Score | 97.8 |
| Open Portuguese LLM - ASSIN2 RTE | 94.61 | Macro F1 (%) | 97.4 |
| Open Korean LLM - Ko-Winogrande | 69.93 | Accuracy (%) | 95.9 |
Interactive version: theaggregate.ai/model?slug=qwen2-5-14b-gutenberg-1e-delta · How It Works · Data refreshed daily, snapshot 2026-10-20.