Mathematical Reasoning on AIME 24 (pass@32, Overall)
56.7Pass@32AV-GRPO
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| AV-GRPOAlgorithm=AV-GRPO, Backbone=Qwen2.5-Math-1.5B, Engine=Gemini-3-pro-preview2026.03 | 56.7 | 50.9 | |
| VM-AV-GRPOAlgorithm=VM-AV-GRPO, Backbone=Qwen2.5-Math-1.5B, Engine=Gemini-3-pro-preview2026.03 | 53.3 | 52.5 | |
| FA-GRPOAlgorithm=FA-GRPO, Backbone=Qwen2.5-Math-1.5B, Engine=Gemini-3-pro-preview2026.03 | 53.3 | 49.9 | |
| PR-GRPOAlgorithm=PR-GRPO, Backbone=Qwen2.5-Math-1.5B, Engine=Gemini-3-pro-preview2026.03 | 53.3 | 49.4 | |
| GRPOAlgorithm=GRPO, Backbone=Qwen2.5-Math-1.5B, Engine=Gemini-3-pro-preview2026.03 | 50 | 47.8 | |
| MSA-GRPOAlgorithm=MSA-GRPO, Backbone=Qwen2.5-Math-1.5B, Engine=Gemini-3-pro-preview2026.03 | 50 | 50.7 | |
| SVE-LNA-GRPOAlgorithm=SVE-LNA-GRPO, Backbone=Qwen2.5-Math-1.5B, Engine=Gemini-3-pro-preview2026.03 | 50 | 48.7 | |
| CDA-GRPOAlgorithm=CDA-GRPO, Backbone=Qwen2.5-Math-1.5B, Engine=Gemini-3-pro-preview2026.03 | 50 | 49 |