Multi-way Classification on 20 system prompts
63SFT AccuracyGPT-4O-MINI
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| GPT-4O-MINIModel=GPT-4O-MINI2026.06 | 63 | 51 | |
| GPT-4.1-NANOModel=GPT-4.1-NANO2026.06 | 59 | 55 | |
| QWEN-4B-INSTRUCTModel=QWEN-4B-INSTRUCT2026.06 | 39 | 37 | |
| LLAMA-3.1-8B-INSTRUCTModel=LLAMA-3.1-8B-INSTRUCT2026.06 | 5 | 21 | |
| GPT-OSS-120BModel=GPT-OSS-120B2026.06 | 5 | 21 | |
| GPT-OSS-20BModel=GPT-OSS-20B2026.06 | 4 | 20 |