List your company

Hallucination Mitigation on ConFiQA

99FaithBALTO
66.96875.28483.691.916Jun 14, 2026

Evaluation results

MethodLinks
BALTOModel=Qwen3-4B2026.069996.295.238.6
GRPOBModel=Qwen3-4B2026.0697.896.394.236.2
BALTOModel=Qwen3-8B2026.0697.596.293.880.2
FSPOModel=Qwen3-8B2026.0694.897.392.266.4
FSPOModel=Qwen3-4B2026.0694.897.392.266.4
GRPODModel=Qwen3-8B2026.0694.595.29074.4
GRPOBModel=Qwen3-8B2026.0692.893.586.873.8
GRPODModel=Qwen3-4B2026.068892.881.737.8
BaseModel=Qwen3-4B2026.0680.3—80.336.8
DPOModel=Qwen3-8B2026.0672.88461.262
SFTModel=Qwen3-8B2026.0671.26848.471.6
BaseModel=Qwen3-8B2026.0668.2—68.270.6