Narrow finetuning on Auto
83.4EM-F1TReFT
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| TReFTModel=Qwen-2.5-32B2026.06 | 83.4 | -0.2 | |
| TReFTModel=Llama-3.1-8B2026.06 | 77.9 | 0.4 | |
| TReFTModel=Qwen-2.5-7B2026.06 | 76.5 | 1.6 | |
| Data interleavingModel=Qwen-2.5-32B2026.06 | 74.9 | -1 | |
| TReFTModel=GPT-OSS-20B2026.06 | 74.3 | -0.81 | |
| Data interleavingModel=Qwen-2.5-7B2026.06 | 70.9 | 0.3 | |
| Data interleavingModel=Llama-3.1-8B2026.06 | 70.2 | 0.2 | |
| SFTModel=Qwen-2.5-7B2026.06 | 68.6 | 0.6 | |
| SFTModel=GPT-OSS-20B2026.06 | 64 | -2.77 | |
| Data interleavingModel=GPT-OSS-20B2026.06 | 63.6 | -1.46 | |
| SFTModel=Llama-3.1-8B2026.06 | 54.2 | -1 | |
| SFTModel=Qwen-2.5-32B2026.06 | 53.4 | -2 |