List your company
ResearchFormerly WizWand

General Reasoning on GPQA-Diamond & MMLU-Pro

57.6AccuracySORT
7.57620.56333.5546.537May 10, 2026May 11, 2026May 12, 2026
Updated 5mo ago

Evaluation results

MethodLinks
SORTBackbone=Qwen3-4B-Instruct2026.0557.6——0.6
Qwen3-4B-InstructBackbone=Qwen3-4B-Instruct2026.0557.6——0
GRPOBackbone=Qwen3-4B-Instruct2026.0557——0.2
ReLIFTBackbone=Qwen3-4B-Instruct2026.0556.8——0
LUFFYBackbone=Qwen3-4B-Instruct2026.0556——-13
Scaf-GRPOBackbone=Qwen3-4B-Instruct2026.0553.6——-1.4
AdaptThinkBackbone=DeepSeek-R1-Distill-Qwen-7B, Trained with same data and RL paradigm=True2026.0548.725,410.610.49—
VPG-EABackbone=DeepSeek-R1-Distill-Qwen-7B, Trained with same data and RL paradigm=True2026.0543.833,617.530.55—
JETBackbone=DeepSeek-R1-Distill-Qwen-7B2026.0543.223,416.070.55—
Original ModelBackbone=DeepSeek-R1-Distill-Qwen-7B2026.0542.25,293.270.38—
R1-VeriThinkerBackbone=DeepSeek-R1-Distill-Qwen-7B2026.0540.364,108.680.4—
AutoThinkBackbone=DeepSeek-R1-Distill-Qwen-7B, Trained with same data and RL paradigm=True2026.0539.814,516.540.36—
SORTBackbone=Qwen2.5-7B-Instruct2026.0539.6——3.4
LaserBackbone=DeepSeek-R1-Distill-Qwen-7B, Trained with same data and RL paradigm=True2026.0538.453,816.830.39—
LUFFYBackbone=Qwen2.5-7B-Instruct2026.0538.1——1.9
GRPOBackbone=Qwen2.5-7B-Instruct2026.0537.2——0.7
Qwen2.5-7B-InstructBackbone=Qwen2.5-7B-Instruct2026.0537.1——0
Scaf-GRPOBackbone=Qwen2.5-7B-Instruct2026.0536.6——0.8
ReLIFTBackbone=Qwen2.5-7B-Instruct2026.0532.4——-1.8
ThinklessBackbone=DeepSeek-R1-Distill-Qwen-1.5B2026.0531.955,403.760.22—
SFTBackbone=Qwen3-4B-Instruct2026.0529.4——-23.3
LaserBackbone=DeepSeek-R1-Distill-Qwen-1.5B, Trained with same data and RL paradigm=True2026.0528.094,468.940.22—
VPG-EABackbone=DeepSeek-R1-Distill-Qwen-1.5B, Trained with same data and RL paradigm=True2026.0528.024,634.170.18—
Original ModelBackbone=DeepSeek-R1-Distill-Qwen-1.5B2026.0527.285,794.190.14—
SORTBackbone=Llama-3.2-3B-Instruct2026.0526.4——11.3
GRPOBackbone=Llama-3.2-3B-Instruct2026.0526.3——10.6
AdaptThinkBackbone=DeepSeek-R1-Distill-Qwen-1.5B, Trained with same data and RL paradigm=True2026.0525.974,064.840.17—
JETBackbone=DeepSeek-R1-Distill-Qwen-1.5B2026.0525.934,534.660.17—
Scaf-GRPOBackbone=Llama-3.2-3B-Instruct2026.0524.1——8.5
AutoThinkBackbone=DeepSeek-R1-Distill-Qwen-1.5B, Trained with same data and RL paradigm=True2026.0523.974,656.370.13—
Llama-3.2-3B-InstructBackbone=Llama-3.2-3B-Instruct2026.0517.9——0
LUFFYBackbone=Llama-3.2-3B-Instruct2026.0516——-1.1
ReLIFTBackbone=Llama-3.2-3B-Instruct2026.0511.7——-3.4
SFTBackbone=Llama-3.2-3B-Instruct2026.0511.6——-7.3
SFTBackbone=Qwen2.5-7B-Instruct2026.059.5——-24.2