News Classification on BBCN (val)
31.8AccuracySFT only
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| SFT onlyModel=Qwen3-4B-Base, N=4, Source Dataset=AGN, OOD=true2025.09 | 31.8 | 56 | |
| ICLModel=Qwen3-4B-Base, N=4, Source Dataset=AGN, OOD=true2025.09 | 28.8 | 12 | |
| IA2 → SFTModel=Qwen3-4B-Base, N=4, Source Dataset=AGN, OOD=true2025.09 | 27.9 | 40 | |
| IA2 onlyModel=Qwen3-4B-Base, N=4, Source Dataset=AGN, OOD=true2025.09 | 24.3 | 10 |