Instruction Following on Air-Bench
90ScoreConfident Decoding
Evaluation Results
| Method | Links | |
|---|---|---|
| Confident DecodingModel=gpt-oss-120B, Decoding Method=Confident2026.06 | 90 | |
| Confident DecodingModel=gpt-oss-20b, Decoding Method=Confident2026.06 | 89 | |
| Last Layer DecodingModel=gpt-oss-120B, Decoding Method=Last Layer2026.06 | 88 | |
| Last Layer DecodingModel=gpt-oss-20b, Decoding Method=Last Layer2026.06 | 86 | |
| Last Layer DecodingModel=Qwen3.5-9B-Base, Decoding Method=Last Layer2026.06 | 76 | |
| Confident DecodingModel=Qwen3.5-9B-Base, Decoding Method=Confident2026.06 | 76 | |
| Confident DecodingModel=Gemma-4-31B, Decoding Method=Confident2026.06 | 76 | |
| Last Layer DecodingModel=Gemma-4-31B, Decoding Method=Last Layer2026.06 | 74 | |
| Last Layer DecodingModel=Qwen3.5-27B, Decoding Method=Last Layer2026.06 | 67 | |
| Confident DecodingModel=Qwen3.5-122B-A10B, Decoding Method=Confident2026.06 | 67 | |
| Last Layer DecodingModel=Qwen3.5-122B-A10B, Decoding Method=Last Layer2026.06 | 66 | |
| Confident DecodingModel=Qwen3.5-27B, Decoding Method=Confident2026.06 | 64 | |
| Confident DecodingModel=Qwen3.5-9B, Decoding Method=Confident2026.06 | 56 | |
| Confident DecodingModel=Qwen3.5-35B-A3B, Decoding Method=Confident2026.06 | 56 | |
| Confident DecodingModel=Qwen3.5-0.8B, Decoding Method=Confident2026.06 | 54 | |
| Last Layer DecodingModel=Qwen3.5-9B, Decoding Method=Last Layer2026.06 | 53 | |
| Last Layer DecodingModel=Qwen3.5-0.8B, Decoding Method=Last Layer2026.06 | 52 | |
| Last Layer DecodingModel=Qwen3.5-35B-A3B, Decoding Method=Last Layer2026.06 | 51 |