List your company

Scientific Reasoning on GPQA-D (Acc, TR)

75.8AccuracyPUMA
17.97632.9884863.012May 17, 2026

Evaluation results

MethodLinks
PUMAModel=Qwen3-30B-A3B-Thinking2026.0575.817.8
DEERModel=Qwen3-30B-A3B-Thinking2026.0574.8-8.2
No-ThinkModel=Qwen3-30B-A3B-Thinking2026.0573.7-11.5
Full CoTModel=Qwen3-30B-A3B-Thinking2026.0572.70
CCoTModel=Qwen3-30B-A3B-Thinking2026.0572.745.5
DynasorModel=Qwen3-30B-A3B-Thinking2026.0567.732.2
CoDModel=Qwen3-30B-A3B-Thinking2026.0565.228.8
Plan&BudgetModel=Qwen3-30B-A3B-Thinking2026.0564.737.1
DEERModel=DeepSeek-R1-Distill-Qwen-7B2026.0551.5-17.4
CoDModel=DeepSeek-R1-Distill-Qwen-7B2026.0550.550.8
DEERModel=Llama-3.1-Nemotron-Nano-8B2026.0549.5-313.2
Full CoTModel=DeepSeek-R1-Distill-Qwen-7B2026.05490
PUMAModel=DeepSeek-R1-Distill-Qwen-7B2026.05494
PUMAModel=Llama-3.1-Nemotron-Nano-8B2026.0548.59.3
CCoTModel=DeepSeek-R1-Distill-Qwen-7B2026.054855.6
Full CoTModel=Llama-3.1-Nemotron-Nano-8B2026.05480
CoDModel=Llama-3.1-Nemotron-Nano-8B2026.054531.9
DynasorModel=Llama-3.1-Nemotron-Nano-8B2026.05451.5
CCoTModel=Llama-3.1-Nemotron-Nano-8B2026.0543.931.1
Plan&BudgetModel=Llama-3.1-Nemotron-Nano-8B2026.0540.926.7
DynasorModel=DeepSeek-R1-Distill-Qwen-7B2026.0539.466.6
Plan&BudgetModel=DeepSeek-R1-Distill-Qwen-7B2026.0538.436.2
Ans. Conv.Model=Qwen3-30B-A3B-Thinking2026.0535.995.5
Ans. Conv.Model=DeepSeek-R1-Distill-Qwen-7B2026.0534.392.4
No-ThinkModel=DeepSeek-R1-Distill-Qwen-7B2026.0528.387.3
No-ThinkModel=Llama-3.1-Nemotron-Nano-8B2026.0525.8-2.2
Ans. Conv.Model=Llama-3.1-Nemotron-Nano-8B2026.0520.295.3