List your company

Graduate-level Reasoning on GPQA-D

49.5AccuracyGeoFaith
22.66829.63436.643.566May 26, 2026

Evaluation results

MethodLinks
GeoFaithBackbone=Qwen3-4B2026.0549.51,60051.3
TruthRLBackbone=Qwen3-4B2026.0544.41,20031.2
THSBackbone=Qwen3-4B2026.0544.41,60032.3
KnowRLBackbone=Qwen3-4B2026.0541.92,00039
GeoFaithBackbone=Qwen3-1.7B2026.0538.490030.3
GRPOBackbone=Qwen3-4B2026.0538.42,50037.7
OriginalBackbone=Qwen3-4B2026.0537.45,50045.7
GRPOBackbone=Qwen3-1.7B2026.0531.31,00029.9
KnowRLBackbone=Qwen3-1.7B2026.0530.880018.5
THSBackbone=Qwen3-1.7B2026.0529.330016
TruthRLBackbone=Qwen3-1.7B2026.0525.86003.6
OriginalBackbone=Qwen3-1.7B2026.0523.71,30027