List your company

Science Question Answering on GPQA Diamond (Acc, throughput)

63.1AccuracyDense
18.58830.14441.753.256Jul 29, 2025

Evaluation results

MethodLinks
DenseModel=DeepSeek R1 Distill Qwen 32B2025.0763.1112.4
DenseModel=QwQ 32B2025.0761.179.06
ReasonCacheModel=DeepSeek R1 Distill Qwen 32B2025.0760.7176.21
ReasonCacheModel=QwQ 32B2025.0758.1125.67
DenseModel=Phi 4 reasoning plus2025.0752175.04
ReasonCacheModel=Phi 4 reasoning plus2025.0750.8243.22
RandomModel=DeepSeek R1 Distill Qwen 32B2025.0732.3172.91
RandomModel=QwQ 32B2025.0727.8123.24
RandomModel=Phi 4 reasoning plus2025.0720.3241.81