List your company

Mathematical Reasoning on AIME 26 (mean@8, pass@8)

70Mean@8DeepSeek-R1-Distill-Qwen-32B
3.720.912538.12555.3375May 29, 2026May 30, 2026

Evaluation results

MethodLinks
DeepSeek-R1-Distill-Qwen-32B2026.057083.33
SkyWork-OR1-Math-7B2026.0562.0880
GRPOStudent Model=DeepSeek-R1-Distill-Qwen-7B, Teacher Model=DeepSeek-R1-Distill-Qwen-32B2026.0562.0880
REOPOLDStudent Model=DeepSeek-R1-Distill-Qwen-7B, Teacher Model=DeepSeek-R1-Distill-Qwen-32B2026.0554.1780
LGRStudent Model=DeepSeek-R1-Distill-Qwen-7B, Teacher Model=DeepSeek-R1-Distill-Qwen-32B2026.0553.7580
LGR (topk)Student Model=DeepSeek-R1-Distill-Qwen-7B, Teacher Model=DeepSeek-R1-Distill-Qwen-32B2026.0553.7576.67
OPD (9k)Student Model=DeepSeek-R1-Distill-Qwen-7B, Teacher Model=DeepSeek-R1-Distill-Qwen-32B2026.0552.576.67
OPD (topk)Student Model=DeepSeek-R1-Distill-Qwen-7B, Teacher Model=DeepSeek-R1-Distill-Qwen-32B2026.0552.0870
JSDStudent Model=DeepSeek-R1-Distill-Qwen-7B, Teacher Model=DeepSeek-R1-Distill-Qwen-32B2026.0551.2566.67
Base ModelStudent Model=DeepSeek-R1-Distill-Qwen-7B2026.0542.9263.33
LGRStudent Model=DeepSeek-R1-Distill-Qwen-1.5B, Teacher Model=SkyWork-OR1-Math-7B2026.0534.1763.33
OPD (topk)Student Model=DeepSeek-R1-Distill-Qwen-1.5B, Teacher Model=SkyWork-OR1-Math-7B2026.0533.7566.67
LGR (topk)Student Model=DeepSeek-R1-Distill-Qwen-1.5B, Teacher Model=SkyWork-OR1-Math-7B2026.0533.7563.33
JSDStudent Model=DeepSeek-R1-Distill-Qwen-1.5B, Teacher Model=SkyWork-OR1-Math-7B2026.0532.550
OPD (9k)Student Model=DeepSeek-R1-Distill-Qwen-1.5B, Teacher Model=SkyWork-OR1-Math-7B2026.0532.0856.67
REOPOLDStudent Model=DeepSeek-R1-Distill-Qwen-1.5B, Teacher Model=SkyWork-OR1-Math-7B2026.053046.67
GRPOStudent Model=DeepSeek-R1-Distill-Qwen-1.5B, Teacher Model=SkyWork-OR1-Math-7B2026.0528.7550
DAPO w/ RRModel Backbone=Qwen3-8B-Base2026.0527.543.33
DAPO w/ PR-RKLModel Backbone=Qwen3-8B-Base2026.0525.8340
DAPOModel Backbone=Qwen3-8B-Base2026.0525.4243.33
DAPO w/ PR-FKLModel Backbone=Qwen3-8B-Base2026.052550
Base ModelStudent Model=DeepSeek-R1-Distill-Qwen-1.5B2026.0517.9246.67
DAPO w/ PR-FKLModel Backbone=Qwen3-4B-Base2026.0517.0830
DAPO w/ PR-RKLModel Backbone=Qwen3-4B-Base2026.0515.8333.33
DAPO w/ RRModel Backbone=Qwen3-4B-Base2026.051530
DAPOModel Backbone=Qwen3-4B-Base2026.0514.1730
BaseModel Backbone=Qwen3-8B-Base2026.056.6720
BaseModel Backbone=Qwen3-4B-Base2026.056.2516.67