List your company
ResearchFormerly WizWand

Math on MATH500 (Score %)

99.32ScoreAIPO
33.051250.255667.4684.6644Apr 17, 2026Apr 21, 2026Apr 26, 2026May 1, 2026May 6, 2026May 11, 2026May 16, 2026
Updated 5mo ago

Evaluation results

MethodLinks
AIPOModel=Qwen2.5-32B-Instruct2026.0599.32
GRPOModel=Qwen2.5-32B-Instruct2026.0598.56
SFTModel=Qwen2.5-32B-Instruct2026.0598.27
OriginalModel=Qwen2.5-32B-Instruct2026.0596.35
Pair+Backbone=olmo-3.1-32b-it-sft2026.0490.8
Mag+Backbone=olmo-3.1-32b-it-sft2026.0490.6
MPO+Backbone=olmo-3.1-32b-it-sft2026.0490
Soft+Backbone=olmo-3.1-32b-it-sft2026.0489.8
Pair+Backbone=olmo-3-7b-it-sft2026.0489.2
AIPOStudent LLM=DeepSeek-R1-Distill-Qwen-1.5B, Teacher LLM=Qwen3-30B-A3B-Instruct-2507, Runs (k)=42026.0588.47
MPO+Backbone=olmo-3-7b-it-sft2026.0488.4
Soft+Backbone=olmo-3-7b-it-sft2026.0487.8
Qwen3-14B + NGMModel Scale=14B, NGM Configuration=True, Decoding Settings=identical decoding settings2026.0587
GRPOStudent LLM=DeepSeek-R1-Distill-Qwen-1.5B, Teacher LLM=Qwen3-30B-A3B-Instruct-2507, Runs (k)=42026.0586.82
Mag+Backbone=olmo-3-7b-it-sft2026.0486.6
DPOBackbone=olmo-3.1-32b-it-sft2026.0486.6
Qwen3-14BModel Scale=14B, NGM Configuration=False, Decoding Settings=identical decoding settings2026.0585.4
Qwen3-8B + NGMModel Scale=8B, NGM Configuration=True, Decoding Settings=identical decoding settings2026.0584.4
Qwen3-4BModel Scale=4B, NGM Configuration=False, Decoding Settings=identical decoding settings2026.0584
OriginalStudent LLM=DeepSeek-R1-Distill-Qwen-1.5B, Teacher LLM=Qwen3-30B-A3B-Instruct-2507, Runs (k)=42026.0583.95
Qwen3-4B + NGMModel Scale=4B, NGM Configuration=True, Decoding Settings=identical decoding settings2026.0583.8
RFTBackbone=olmo-3.1-32b-it-sft2026.0483
Qwen3-8BModel Scale=8B, NGM Configuration=False, Decoding Settings=identical decoding settings2026.0583
DPOBackbone=olmo-3-7b-it-sft2026.0481
RFTBackbone=olmo-3-7b-it-sft2026.0477.6
SFTStudent LLM=DeepSeek-R1-Distill-Qwen-1.5B, Teacher LLM=Qwen3-30B-A3B-Instruct-2507, Runs (k)=42026.0577.16
BaseBackbone=olmo-3.1-32b-it-sft2026.0476.4
Qwen3-1.7B + NGMModel Scale=1.7B, NGM Configuration=True, Decoding Settings=identical decoding settings2026.0574
Qwen3-1.7BModel Scale=1.7B, NGM Configuration=False, Decoding Settings=identical decoding settings2026.0571.4
BaseBackbone=olmo-3-7b-it-sft2026.0469.4
Qwen3-0.6BModel Scale=0.6B, NGM Configuration=False, Decoding Settings=identical decoding settings2026.0551.4
Qwen3-0.6B + NGMModel Scale=0.6B, NGM Configuration=True, Decoding Settings=identical decoding settings2026.0549.8
Mag+Backbone=gemma-3-4b-sft2026.0440.4
DPOBackbone=gemma-3-4b-sft2026.0440.2
Soft+Backbone=gemma-3-4b-sft2026.0440.2
Pair+Backbone=gemma-3-4b-sft2026.0440
MPO+Backbone=gemma-3-4b-sft2026.0439.6
RFTBackbone=gemma-3-4b-sft2026.0438.4
BaseBackbone=gemma-3-4b-sft2026.0435.6