Question Answering on SQuAD (Faith, Acc.)
0.9463FaithfulnessFaithRL
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| FaithRLModel=Qwen3-1.7B2026.02 | 0.9463 | 0.7527 | |
| Self-RefineModel=Qwen3-1.7B2026.02 | 0.942 | 0.7415 | |
| SFTModel=Qwen3-1.7B2026.02 | 0.9384 | 0.7278 | |
| KDModel=Qwen3-1.7B2026.02 | 0.9381 | 0.7156 | |
| FSPOModel=Qwen3-1.7B2026.02 | 0.9357 | 0.7183 | |
| BaseModel=Qwen3-1.7B2026.02 | 0.934 | 0.7348 | |
| FaithRLModel=Qwen3-0.6B2026.02 | 0.8924 | 0.6918 | |
| GRPOModel=Qwen3-1.7B2026.02 | 0.8895 | 0.7404 | |
| FaithRLModel=DPSK-1.5B2026.02 | 0.878 | 0.6332 | |
| Self-RefineModel=Qwen3-0.6B2026.02 | 0.876 | 0.6693 | |
| SFTModel=Qwen3-0.6B2026.02 | 0.8731 | 0.6512 | |
| BaseModel=Qwen3-0.6B2026.02 | 0.866 | 0.6359 | |
| KDModel=Qwen3-0.6B2026.02 | 0.8642 | 0.6512 | |
| GRPOModel=DPSK-1.5B2026.02 | 0.856 | 0.4378 | |
| FSPOModel=Qwen3-0.6B2026.02 | 0.8517 | 0.6534 | |
| GRPOModel=Qwen3-0.6B2026.02 | 0.82 | 0.5048 | |
| KDModel=DPSK-1.5B2026.02 | 0.812 | 0.5498 | |
| SFTModel=DPSK-1.5B2026.02 | 0.7526 | 0.442 | |
| Self-RefineModel=DPSK-1.5B2026.02 | 0.7374 | 0.4134 | |
| BaseModel=DPSK-1.5B2026.02 | 0.734 | 0.4349 | |
| FSPOModel=DPSK-1.5B2026.02 | 0.7283 | 0.5646 |