Mathematical Reasoning on Math Avg
47.6AccuracyMargin+
Evaluation Results
| Method | Links | |
|---|---|---|
| Margin+Backbone=qwen3-4b-base, nll coefficient=1.0, Setting=Online, Checkpoint selection=Highest checkpoint2026.04 | 47.6 | |
| MPO+Backbone=qwen3-4b-base, nll coefficient=1.0, Setting=Online, Checkpoint selection=Highest checkpoint2026.04 | 47.6 | |
| Softmax+Backbone=qwen3-4b-base, nll coefficient=1.0, Setting=Online, Checkpoint selection=Highest checkpoint2026.04 | 47.3 | |
| All-Pairs+Backbone=qwen3-4b-base, nll coefficient=1.0, Setting=Online, Checkpoint selection=Highest checkpoint2026.04 | 46.6 | |
| Margin+Backbone=qwen3-4b-base, nll coefficient=0.0, Setting=Online, Checkpoint selection=Highest checkpoint2026.04 | 46.6 | |
| All-Pairs+Backbone=qwen3-4b-base, nll coefficient=0.0, Setting=Online, Checkpoint selection=Highest checkpoint2026.04 | 45.9 | |
| DPOBackbone=qwen3-4b-base, nll coefficient=1.0, Setting=Online, Checkpoint selection=Highest checkpoint2026.04 | 45.7 | |
| MPO+Backbone=qwen3-4b-base, nll coefficient=0.0, Setting=Online, Checkpoint selection=Highest checkpoint2026.04 | 45.6 | |
| Softmax+Backbone=qwen3-4b-base, nll coefficient=0.0, Setting=Online, Checkpoint selection=Highest checkpoint2026.04 | 44.1 | |
| DPOBackbone=qwen3-4b-base, nll coefficient=0.0, Setting=Online, Checkpoint selection=Highest checkpoint2026.04 | 43.5 | |
| RFTBackbone=qwen3-4b-base, nll coefficient=0.0, Setting=Online, Checkpoint selection=Highest checkpoint2026.04 | 43.1 |