List your company

Scientific Reasoning on GPQA (Acc, Time, Peak, Dep)

40.91AccuracyVanilla
-1.63649.409320.45531.5007Apr 4, 2026

Evaluation results

MethodLinks
VanillaBackbone=Llama3.1-8B2026.0440.917.986,62226
LThinker_thoBackbone=Llama3.1-8B2026.0436.366.381,7966.4
LThinker_tokBackbone=Llama3.1-8B2026.0431.826.943,15012
Vanilla + H2OBackbone=Llama3.1-8B2026.0431.8112.551,5367.9
AnLLMBackbone=Llama3.1-8B2026.0431.317.198383.7
VanillaBackbone=Qwen2.5-7B2026.0430.8110.768,05539
LThinker_thoBackbone=Qwen2.5-7B2026.0430.38.412,3859.3
LThinker_tokBackbone=Qwen2.5-7B2026.0428.288.263,94018
CoTBackbone=Qwen2.5-7B2026.0426.760.69680.5
Distill-R1Backbone=Qwen2.5-7B2026.0424.758.016,71831
Vanilla + H2OBackbone=Qwen2.5-7B2026.0424.7515.611,2009.8
CoTBackbone=Llama3.1-8B2026.0424.750.961,2310.9
Distill-R1Backbone=Llama3.1-8B2026.0420.25.243,97216
AnLLMBackbone=Qwen2.5-7B2026.0419.79.143,40111
Vanilla + SepLLMBackbone=Llama3.1-8B2026.042.5312.621,02410
Vanilla + SepLLMBackbone=Qwen2.5-7B2026.04011.651,02410