List your company
ResearchFormerly WizWand

Scientific Reasoning on GPQA Diamond (pass@1)

69.5pass@1SPLA
8.76424.53240.356.068Oct 2, 2025Oct 21, 2025Nov 10, 2025Nov 30, 2025Dec 20, 2025Jan 9, 2026Jan 29, 2026
Updated 8mo ago

Evaluation results

MethodLinks
SPLAModel Size=14B, Temperature=0.6, Max output length=32k, Training context length=32k, Training tokens=1.2T, Sampling strategy=pass@12026.0169.5
SPAModel Size=14B, Temperature=0.6, Max output length=32k, Training context length=32k, Training tokens=1.2T, Sampling strategy=pass@12026.0169.2
InfLLM-v2Model Size=14B, Temperature=0.6, Max output length=32k, Training context length=32k, Training tokens=1.2T, Sampling strategy=pass@12026.0168.7
Dense AttentionModel Size=14B, Temperature=0.6, Max output length=32k, Training context length=32k, Training tokens=1.2T, Sampling strategy=pass@12026.0168.5
NSAModel Size=14B, Temperature=0.6, Max output length=32k, Training context length=32k, Training tokens=1.2T, Sampling strategy=pass@12026.0159.6
Continual LUFFYBackbone=Qwen2.5-Math-7B, RLVR Setup=Continual RLVR2025.1049
On-Policy (Continual)Backbone=Qwen2.5-Math-7B, RLVR Setup=Continual RLVR2025.1047
ExGRPO (Continual)Backbone=Qwen2.5-Math-7B, RLVR Setup=Continual RLVR2025.1042.4
GPG-ZeroBackbone=Qwen2.5-Math-7B, RLVR Setup=Previous Zero RLVR2025.1040.4
LUFFYBackbone=Qwen2.5-Math-7B, RLVR Setup=Continual RLVR2025.1039.9
On-PolicyBackbone=Qwen2.5-Math-7B, RLVR Setup=Zero RLVR2025.1037.4
ExGRPOBackbone=Qwen2.5-Math-7B, RLVR Setup=Zero RLVR2025.1037.4
Qwen-InstructBackbone=Qwen2.5-Math-7B2025.1024.7
SFTBackbone=Qwen2.5-Math-7B, RLVR Setup=Off-policy Learning2025.1024.7
RePO-ZeroBackbone=Qwen2.5-Math-7B, RLVR Setup=Previous Zero RLVR2025.1024.2
SFT+RLBackbone=Qwen2.5-Math-7B, RLVR Setup=Off-policy Learning2025.1024.2
Oat-ZeroBackbone=Qwen2.5-Math-7B, RLVR Setup=Previous Zero RLVR2025.1023.7
PRIME-ZeroBackbone=Qwen2.5-Math-7B, RLVR Setup=Previous Zero RLVR2025.1018.2
Qwen-BaseBackbone=Qwen2.5-Math-7B2025.1011.1