Yi-1.5-9B-Chat-abliterated: benchmark results
byroneverson's abliterated Yi-1.5 9B Chat with the refusal direction removed, with the process documented in an included Jupyter notebook. Provider: byroneverson. Released 2024-09-04. Access: Open.
Aggregate Rating 1420 ± 18, rank #1149 of 1656 rated models, from 29 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Open Korean LLM - Ko-IFEval | 40.8 | Strict accuracy, prompt/instruction mean (%) | 83.6 |
| Open Chinese LLM - GSM8K | 59.14 | Accuracy (%) | 83.1 |
| Open LLM Leaderboard - MuSR | 43.89 | Score | 77.4 |
| Open Korean LLM - Ko-EQ-Bench | 28.16 | EQ-Bench score | 75.5 |
| Open Chinese LLM Leaderboard | 62.94 | Official Nonzero-Task Mean (%) | 75.4 |
| Open Portuguese LLM - FaQuAD NLI | 75.29 | Macro F1 (%) | 73.2 |
| Open Chinese LLM - C-Eval Semantic | 82.13 | Accuracy (%) | 72.2 |
| Open Chinese LLM - ARC Challenge | 55.63 | Accuracy (%) | 69.2 |
| Open LLM Leaderboard - IFEval | 57.23 | Score | 69.1 |
| Open LLM Leaderboard - BBH | 54.01 | Score | 69 |
| Open Chinese LLM - CMMLU | 59.04 | Accuracy (%) | 66.9 |
| Open LLM Leaderboard - MATH Level 5 | 16.62 | Score | 64.9 |
Interactive version: theaggregate.ai/model?slug=yi-1-5-9b-chat-abliterated · How It Works · Data refreshed daily, snapshot 2026-10-20.