Narrow finetuning on Health
83.9EM-F1Data interleaving
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Data interleavingModel=Qwen-2.5-32B2026.06 | 83.9 | -0.5 | |
| TReFTModel=Qwen-2.5-32B2026.06 | 83 | -0.3 | |
| TReFTModel=Llama-3.1-8B2026.06 | 80.1 | 0.5 | |
| TReFTModel=Qwen-2.5-7B2026.06 | 79.9 | 0.2 | |
| Data interleavingModel=Qwen-2.5-7B2026.06 | 76.2 | -0.4 | |
| Data interleavingModel=Llama-3.1-8B2026.06 | 74.5 | 0.2 | |
| TReFTModel=GPT-OSS-20B2026.06 | 69.6 | -1.58 | |
| SFTModel=Qwen-2.5-32B2026.06 | 61.1 | -1.3 | |
| SFTModel=GPT-OSS-20B2026.06 | 54 | -3.21 | |
| SFTModel=Llama-3.1-8B2026.06 | 53.4 | 0.1 | |
| Data interleavingModel=GPT-OSS-20B2026.06 | 53.4 | -1.74 | |
| SFTModel=Qwen-2.5-7B2026.06 | 44.7 | -1.8 |