List your company

Scientific Reasoning on GPQA (avg@16)

44.1avg@16Qwen2.5-32B-Instruct
25.06830.00934.9539.891Mar 19, 2026

Evaluation results

MethodLinks
Qwen2.5-32B-InstructBackbone=Qwen2.5-32B-Instruct2026.0344.1
PowerFlowBackbone=Qwen2.5-Math-7B2026.0337
PowerSamplingBackbone=Qwen2.5-Math-7B2026.0336.3
EMPOBackbone=Qwen2.5-Math-7B2026.0336
InstructBackbone=Qwen2.5-Math-7B2026.0335
TTRLBackbone=Qwen2.5-Math-7B2026.0334.7
GRPOBackbone=Qwen2.5-Math-7B2026.0334.4
Low-tempBackbone=Qwen2.5-Math-7B2026.0334.1
One-shot EMBackbone=Qwen2.5-Math-7B2026.0332.8
IntuitorBackbone=Llama-3.2-3B-Instruct2026.0330.5
GRPOBackbone=Llama-3.2-3B-Instruct2026.0330.5
PowerFlowBackbone=Llama-3.2-3B-Instruct2026.0330.2
BaseBackbone=Qwen2.5-Math-7B2026.0329.7
EMPOBackbone=Qwen2.5-Math-1.5B2026.0329.5
BaseBackbone=Llama-3.2-3B-Instruct2026.0329.5
Low-tempBackbone=Llama-3.2-3B-Instruct2026.0328.8
Low-tempBackbone=Qwen2.5-Math-1.5B2026.0328.7
PowerFlowBackbone=Qwen2.5-Math-1.5B2026.0328.3
InstructBackbone=Qwen2.5-1.5B2026.0328
PowerFlowBackbone=Qwen2.5-1.5B2026.0327.7
GRPOBackbone=Qwen2.5-Math-1.5B2026.0326.8
Low-tempBackbone=Qwen2.5-1.5B2026.0326.6
IntuitorBackbone=Qwen2.5-1.5B2026.0326.4
InstructBackbone=Qwen2.5-Math-1.5B2026.0326.4
BaseBackbone=Qwen2.5-Math-1.5B2026.0326.1
Format-onlyBackbone=Qwen2.5-Math-1.5B2026.0326.1
GRPOBackbone=Qwen2.5-1.5B2026.0326
BaseBackbone=Qwen2.5-1.5B2026.0325.8