Mathematical Reasoning on MATH500 (pass@k metrics)
91.3Pass@1PieceHint-Nemotron-1.5B
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| PieceHint-Nemotron-1.5BModel Name=Nemotron-1.5B, Model Size=1.5B, Method Type=PieceHint2026.04 | 91.3 | — | |
| PieceHint-Qwen3-1.7BModel Name=Qwen3-1.7B, Model Size=1.7B, Method Type=PieceHint2026.04 | 89.8 | — | |
| Nemotron-1.5BModel Name=Nemotron-1.5B, Model Size=1.5B, Method Type=Baseline2026.04 | 88.5 | — | |
| PieceHint-DeepSeek-1.5BModel Name=DeepSeek-R1-Distill-1.5B, Model Size=1.5B, Method Type=PieceHint2026.04 | 87.7 | — | |
| DeepSeek-R1-Distill-32BModel Name=DeepSeek-R1-Distill-32B, Model Size=32B, Method Type=Baseline2026.04 | 86 | — | |
| Qwen3-4BModel Name=Qwen3-4B, Model Size=4B, Method Type=Baseline2026.04 | 85.1 | — | |
| Qwen3-1.7BModel Name=Qwen3-1.7B, Model Size=1.7B, Method Type=Baseline2026.04 | 84.1 | — | |
| OPOBackbone=Qwen2.5-Math-7B2025.10 | 82.2 | — | |
| STEERBackbone=Qwen2.5-Math-7B2025.10 | 82.2 | — | |
| DeepSeek-R1-Distill-7BModel Name=DeepSeek-R1-Distill-7B, Model Size=7B, Method Type=Baseline2026.04 | 80.3 | — | |
| GRPO w/ Entro. LossBackbone=Qwen2.5-Math-7B2025.10 | 80 | — | |
| Eurus-PRIMEBackbone=Qwen2.5-Math-7B2025.10 | 79.8 | — | |
| Entro. Adv.Backbone=Qwen2.5-Math-7B2025.10 | 79.6 | — | |
| SimpleRL-ZooBackbone=Qwen2.5-Math-7B2025.10 | 79.2 | — | |
| GRPO w/ Fork TokensBackbone=Qwen2.5-Math-7B2025.10 | 79.2 | — | |
| W-REINFORCEBackbone=Qwen2.5-Math-7B2025.10 | 79.2 | — | |
| F-DAPOBackbone=Qwen2.5-7B, Group Size (N)=82026.02 | 79.1 | 96.6 | |
| F-CISPOBackbone=Qwen2.5-7B, Group Size (N)=82026.02 | 79 | 97.8 | |
| GRPO w/ clip-highBackbone=Qwen2.5-Math-7B2025.10 | 79 | — | |
| KL–CovBackbone=Qwen2.5-Math-7B2025.10 | 78.8 | — | |
| CISPOBackbone=Qwen2.5-7B, Group Size (N)=82026.02 | 78.7 | 97 | |
| DAPOBackbone=Qwen2.5-7B, Group Size (N)=82026.02 | 78.6 | 95.2 | |
| GRPOBackbone=Qwen2.5-Math-7B2025.10 | 78.6 | — | |
| Clip–CovBackbone=Qwen2.5-Math-7B2025.10 | 78 | — | |
| CISPOBackbone=Qwen2.5-1.5B-Math, Group Size (N)=82026.02 | 77.3 | 95.5 | |
| F-DAPOBackbone=Qwen2.5-1.5B-Math, Group Size (N)=82026.02 | 76.8 | 97 | |
| F-CISPOBackbone=Qwen2.5-1.5B-Math, Group Size (N)=82026.02 | 76.8 | 96.5 | |
| DAPOBackbone=Qwen2.5-1.5B-Math, Group Size (N)=82026.02 | 76.5 | 96.2 | |
| F-GRPOBackbone=Qwen2.5-7B, Group Size (N)=82026.02 | 76.2 | 95.1 | |
| GRPOBackbone=Qwen2.5-7B, Group Size (N)=82026.02 | 75.8 | 92.8 | |
| GRPOBackbone=Qwen2.5-1.5B-Math, Group Size (N)=82026.02 | 75.4 | 95.6 | |
| F-GRPOBackbone=Qwen2.5-1.5B-Math, Group Size (N)=82026.02 | 74.7 | 96.1 | |
| DeepSeek-R1-Distill-1.5BModel Name=DeepSeek-R1-Distill-1.5B, Model Size=1.5B, Method Type=Baseline2026.04 | 72.1 | — | |
| F-DAPOBackbone=Llama-3.2-3B-Instruct, Group Size (N)=82026.02 | 58.6 | 92 | |
| CISPOBackbone=Llama-3.2-3B-Instruct, Group Size (N)=82026.02 | 56.9 | 89.1 | |
| F-CISPOBackbone=Llama-3.2-3B-Instruct, Group Size (N)=82026.02 | 56.5 | 91 | |
| DAPOBackbone=Llama-3.2-3B-Instruct, Group Size (N)=82026.02 | 55.9 | 83.8 | |
| GRPOBackbone=Llama-3.2-3B-Instruct, Group Size (N)=82026.02 | 55 | 90.6 | |
| F-GRPOBackbone=Llama-3.2-3B-Instruct, Group Size (N)=82026.02 | 54.1 | 92.9 | |
| Qwen2.5-Math-7BBackbone=Qwen2.5-Math-7B2025.10 | 39.6 | — |