Logical Reasoning on Sudoku (Token Length Specific)
23.88Accuracy (128 tokens)AIS
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| AISBackbone=LLaDA-8B, Evaluation Protocol=Full-model fine-tuning2026.05 | 23.88 | 25.07 | 24.07 | |
| FlashRL (TIS)Backbone=LLaDA-8B, Evaluation Protocol=Full-model fine-tuning2026.05 | 23.8 | 24.69 | 23.73 | |
| BF16Backbone=LLaDA-8B, Evaluation Protocol=Full-model fine-tuning2026.05 | 22.1 | 26.7 | 29.25 | |
| FP8 RolloutBackbone=LLaDA-8B, Evaluation Protocol=Full-model fine-tuning2026.05 | 21.63 | 23.3 | 23.01 |