List your company
ResearchFormerly WizWand

Science on MMLU-Pro

90.1AccuracyGemini-3-pro High
-2.4621.5745.669.63May 6, 2026May 15, 2026May 24, 2026Jun 2, 2026Jun 11, 2026Jun 20, 2026Jun 30, 2026
Updated 4mo ago

Evaluation results

MethodLinks
Gemini-3-pro High2026.0690.1
Claude-Opus-4.52026.0689.3
Claude-Sonnet-4.52026.0688
Seed2.0 Pro2026.0687
GPT-5.2 High2026.0685.9
BaseBackbone=Qwen2.5-3B-Instruct2026.0541.9
BaseBackbone=Qwen2.5-3B-Instruct, Incremental Learning Stage=Pre-trained2026.0541.9
Self-SFTBackbone=Qwen2.5-3B-Instruct2026.0538.6
Low-SFTBackbone=Llama-3.2-3B-Instruct2026.0537.8
Iter-SFTBackbone=Qwen2.5-3B-Instruct2026.0537.7
Anchored LearningBackbone=Qwen2.5-3B-Instruct2026.0537.7
Anchored LearningBackbone=Qwen2.5-3B-Instruct, Incremental Learning Stage=Stage 1: iGSM2026.0537.7
Anchored LearningBackbone=Llama-3.2-3B-Instruct2026.0536.8
Low-SFTBackbone=Qwen2.5-3B-Instruct, Incremental Learning Stage=Stage 3: iGSM -> MedCalc -> IFEval2026.0536.8
Anchored LearningBackbone=Qwen2.5-3B-Instruct, Incremental Learning Stage=Stage 3: iGSM -> MedCalc -> IFEval2026.0536.6
Low-SFTBackbone=Llama-3.2-3B-Instruct2026.0536.3
KL-SFTBackbone=Llama-3.2-3B-Instruct2026.0535.5
BaseBackbone=Llama-3.2-3B-Instruct2026.0534.7
BaseBackbone=Llama-3.2-3B-Instruct2026.0534.7
BaseBackbone=Llama-3.2-3B-Instruct2026.0534.7
DFTBackbone=Llama-3.2-3B-Instruct2026.0534.7
Anchored LearningBackbone=Llama-3.2-3B-Instruct2026.0534.3
Anchored LearningBackbone=Llama-3.2-3B-Instruct2026.0533.3
Low-SFTBackbone=Qwen2.5-3B-Instruct2026.0533.3
STMBackbone=Llama-3.2-3B-Instruct2026.0533.3
Low-SFTBackbone=Qwen2.5-3B-Instruct, Incremental Learning Stage=Stage 1: iGSM2026.0533.3
Low-SFTBackbone=Llama-3.2-3B-Instruct2026.0533.2
SFTBackbone=Llama-3.2-3B-Instruct2026.0532.7
Iter-SFTBackbone=Llama-3.2-3B-Instruct2026.0532.2
STMBackbone=Qwen2.5-3B-Instruct2026.0532.1
Self-SFTBackbone=Llama-3.2-3B-Instruct2026.0531.2
Iter-SFTBackbone=Llama-3.2-3B-Instruct2026.0530.9
Anchored LearningBackbone=Qwen2.5-3B-Instruct, Incremental Learning Stage=Stage 2: iGSM -> MedCalc2026.0530.8
Low-SFTBackbone=Qwen2.5-3B-Instruct, Incremental Learning Stage=Stage 2: iGSM -> MedCalc2026.0529.9
Self-SFTBackbone=Llama-3.2-3B-Instruct2026.0529.6
STMBackbone=Llama-3.2-3B-Instruct2026.0529.2
DFTBackbone=Llama-3.2-3B-Instruct2026.0528
Iter-SFTBackbone=Llama-3.2-3B-Instruct2026.0528
Self-SFTBackbone=Llama-3.2-3B-Instruct2026.0526.9
STMBackbone=Llama-3.2-3B-Instruct2026.0524.3
DFTBackbone=Qwen2.5-3B-Instruct2026.0516.3
KL-SFTBackbone=Qwen2.5-3B-Instruct2026.0510.2
KL-SFT 0.2Backbone=Llama-3.2-3B-Instruct2026.059.6
DFTBackbone=Llama-3.2-3B-Instruct2026.058.2
SFTBackbone=Qwen2.5-3B-Instruct2026.053.1
KL-SFTBackbone=Llama-3.2-3B-Instruct2026.052
SFTBackbone=Llama-3.2-3B-Instruct2026.051.2
SFTBackbone=Llama-3.2-3B-Instruct2026.051.1