List your company

Question Answering on HotpotQA (Faith and Accuracy)

93.4FaithfulnessFaithRL
65.52872.7648087.236Feb 5, 2026

Evaluation results

MethodLinks
FaithRLModel=Qwen3-1.7B2026.0293.476.98
KDModel=Qwen3-1.7B2026.0292.5773.1
FSPOModel=Qwen3-1.7B2026.0291.8375.46
SFTModel=Qwen3-1.7B2026.0291.4673.74
Self-RefineModel=Qwen3-1.7B2026.0290.6172.46
BaseModel=Qwen3-1.7B2026.0290.472.81
FaithRLModel=Qwen3-0.6B2026.0286.466.46
GRPOModel=Qwen3-1.7B2026.0285.674.56
FaithRLModel=DPSK-1.5B2026.0285.462.74
FSPOModel=Qwen3-0.6B2026.0284.6163.87
KDModel=Qwen3-0.6B2026.0283.6465.42
SFTModel=Qwen3-0.6B2026.0283.2965.35
GRPOModel=DPSK-1.5B2026.0281.840.12
BaseModel=Qwen3-0.6B2026.0281.663.21
Self-RefineModel=Qwen3-0.6B2026.0280.7964.27
GRPOModel=Qwen3-0.6B2026.0279.8535.6
KDModel=DPSK-1.5B2026.0276.657.46
SFTModel=DPSK-1.5B2026.0268.6552.4
Self-RefineModel=DPSK-1.5B2026.0268.2748.65
FSPOModel=DPSK-1.5B2026.0267.6557.3
BaseModel=DPSK-1.5B2026.0266.649.2