Math on MATH500 (Score %)
99.32ScoreAIPO
Evaluation Results
| Method | Links | |
|---|---|---|
| AIPOModel=Qwen2.5-32B-Instruct2026.05 | 99.32 | |
| GRPOModel=Qwen2.5-32B-Instruct2026.05 | 98.56 | |
| SFTModel=Qwen2.5-32B-Instruct2026.05 | 98.27 | |
| OriginalModel=Qwen2.5-32B-Instruct2026.05 | 96.35 | |
| Pair+Backbone=olmo-3.1-32b-it-sft2026.04 | 90.8 | |
| Mag+Backbone=olmo-3.1-32b-it-sft2026.04 | 90.6 | |
| MPO+Backbone=olmo-3.1-32b-it-sft2026.04 | 90 | |
| Soft+Backbone=olmo-3.1-32b-it-sft2026.04 | 89.8 | |
| Pair+Backbone=olmo-3-7b-it-sft2026.04 | 89.2 | |
| AIPOStudent LLM=DeepSeek-R1-Distill-Qwen-1.5B, Teacher LLM=Qwen3-30B-A3B-Instruct-2507, Runs (k)=42026.05 | 88.47 | |
| MPO+Backbone=olmo-3-7b-it-sft2026.04 | 88.4 | |
| Soft+Backbone=olmo-3-7b-it-sft2026.04 | 87.8 | |
| Qwen3-14B + NGMModel Scale=14B, NGM Configuration=True, Decoding Settings=identical decoding settings2026.05 | 87 | |
| GRPOStudent LLM=DeepSeek-R1-Distill-Qwen-1.5B, Teacher LLM=Qwen3-30B-A3B-Instruct-2507, Runs (k)=42026.05 | 86.82 | |
| Mag+Backbone=olmo-3-7b-it-sft2026.04 | 86.6 | |
| DPOBackbone=olmo-3.1-32b-it-sft2026.04 | 86.6 | |
| Qwen3-14BModel Scale=14B, NGM Configuration=False, Decoding Settings=identical decoding settings2026.05 | 85.4 | |
| Qwen3-8B + NGMModel Scale=8B, NGM Configuration=True, Decoding Settings=identical decoding settings2026.05 | 84.4 | |
| Qwen3-4BModel Scale=4B, NGM Configuration=False, Decoding Settings=identical decoding settings2026.05 | 84 | |
| OriginalStudent LLM=DeepSeek-R1-Distill-Qwen-1.5B, Teacher LLM=Qwen3-30B-A3B-Instruct-2507, Runs (k)=42026.05 | 83.95 | |
| Qwen3-4B + NGMModel Scale=4B, NGM Configuration=True, Decoding Settings=identical decoding settings2026.05 | 83.8 | |
| RFTBackbone=olmo-3.1-32b-it-sft2026.04 | 83 | |
| Qwen3-8BModel Scale=8B, NGM Configuration=False, Decoding Settings=identical decoding settings2026.05 | 83 | |
| DPOBackbone=olmo-3-7b-it-sft2026.04 | 81 | |
| RFTBackbone=olmo-3-7b-it-sft2026.04 | 77.6 | |
| SFTStudent LLM=DeepSeek-R1-Distill-Qwen-1.5B, Teacher LLM=Qwen3-30B-A3B-Instruct-2507, Runs (k)=42026.05 | 77.16 | |
| BaseBackbone=olmo-3.1-32b-it-sft2026.04 | 76.4 | |
| Qwen3-1.7B + NGMModel Scale=1.7B, NGM Configuration=True, Decoding Settings=identical decoding settings2026.05 | 74 | |
| Qwen3-1.7BModel Scale=1.7B, NGM Configuration=False, Decoding Settings=identical decoding settings2026.05 | 71.4 | |
| BaseBackbone=olmo-3-7b-it-sft2026.04 | 69.4 | |
| Qwen3-0.6BModel Scale=0.6B, NGM Configuration=False, Decoding Settings=identical decoding settings2026.05 | 51.4 | |
| Qwen3-0.6B + NGMModel Scale=0.6B, NGM Configuration=True, Decoding Settings=identical decoding settings2026.05 | 49.8 | |
| Mag+Backbone=gemma-3-4b-sft2026.04 | 40.4 | |
| DPOBackbone=gemma-3-4b-sft2026.04 | 40.2 | |
| Soft+Backbone=gemma-3-4b-sft2026.04 | 40.2 | |
| Pair+Backbone=gemma-3-4b-sft2026.04 | 40 | |
| MPO+Backbone=gemma-3-4b-sft2026.04 | 39.6 | |
| RFTBackbone=gemma-3-4b-sft2026.04 | 38.4 | |
| BaseBackbone=gemma-3-4b-sft2026.04 | 35.6 |