List your company

Question Answering on SQuAD (Faith, Acc.)

0.9463FaithfulnessFaithRL
0.719580.778440.83730.89616Feb 5, 2026

Evaluation results

MethodLinks
FaithRLModel=Qwen3-1.7B2026.020.94630.7527
Self-RefineModel=Qwen3-1.7B2026.020.9420.7415
SFTModel=Qwen3-1.7B2026.020.93840.7278
KDModel=Qwen3-1.7B2026.020.93810.7156
FSPOModel=Qwen3-1.7B2026.020.93570.7183
BaseModel=Qwen3-1.7B2026.020.9340.7348
FaithRLModel=Qwen3-0.6B2026.020.89240.6918
GRPOModel=Qwen3-1.7B2026.020.88950.7404
FaithRLModel=DPSK-1.5B2026.020.8780.6332
Self-RefineModel=Qwen3-0.6B2026.020.8760.6693
SFTModel=Qwen3-0.6B2026.020.87310.6512
BaseModel=Qwen3-0.6B2026.020.8660.6359
KDModel=Qwen3-0.6B2026.020.86420.6512
GRPOModel=DPSK-1.5B2026.020.8560.4378
FSPOModel=Qwen3-0.6B2026.020.85170.6534
GRPOModel=Qwen3-0.6B2026.020.820.5048
KDModel=DPSK-1.5B2026.020.8120.5498
SFTModel=DPSK-1.5B2026.020.75260.442
Self-RefineModel=DPSK-1.5B2026.020.73740.4134
BaseModel=DPSK-1.5B2026.020.7340.4349
FSPOModel=DPSK-1.5B2026.020.72830.5646