List your company

Scientific Reasoning on GPQA Diamond (Acc, Len, CR)

63.1AccuracyASAG
23.78833.99444.254.406Jun 13, 2026

Evaluation results

MethodLinks
ASAGBackbone=Qwen3-14B2026.0663.15,07566
DEERBackbone=Qwen3-14B2026.0660.64,99865
DynasorBackbone=Qwen3-14B2026.0659.65,38770.1
TALEBackbone=Qwen3-14B2026.0659.15,39170.1
VanillaBackbone=Qwen3-14B2026.0658.67,688100
ASAGBackbone=Qwen3-8B2026.0656.15,71460.9
DynasorBackbone=Qwen3-8B2026.0655.15,64760.2
DEERBackbone=Qwen3-8B2026.0654.55,33456.9
VanillaBackbone=Qwen3-8B2026.0652.59,382100
TALEBackbone=Qwen3-8B2026.06524,16144.4
NoThinkingBackbone=Qwen3-14B2026.0650.52,30830
NoThinkingBackbone=Qwen3-8B2026.0648.52,56427.3
ASAGBackbone=Qwen3-4B2026.06483,56338.4
DEERBackbone=Qwen3-4B2026.06473,80241
VanillaBackbone=Qwen3-4B2026.0646.59,277100
DynasorBackbone=Qwen3-4B2026.0646.54,17945
TALEBackbone=Qwen3-4B2026.0641.92,62028.2
NoThinkingBackbone=Qwen3-4B2026.0636.41,73818.7
ASAGBackbone=DeepSeek-R1-Distill-Qwen-7B2026.0631.85,58456.8
DynasorBackbone=DeepSeek-R1-Distill-Qwen-7B2026.0629.36,81769.3
DEERBackbone=DeepSeek-R1-Distill-Qwen-7B2026.0628.85,63957.3
NoThinkingBackbone=DeepSeek-R1-Distill-Qwen-7B2026.0626.81,38514.1
TALEBackbone=DeepSeek-R1-Distill-Qwen-7B2026.0626.34,88349.7
VanillaBackbone=DeepSeek-R1-Distill-Qwen-7B2026.0625.39,834100