Mathematical Reasoning on GSM8K (Accuracy @ 128/256/512 Tokens)
74.76Accuracy (128 tokens)AIS
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| AISBackbone=LLaDA-8B, Evaluation Protocol=Full-model fine-tuning2026.05 | 74.76 | 80.52 | 81.94 | |
| BF16Backbone=LLaDA-8B, Evaluation Protocol=Full-model fine-tuning2026.05 | 74.6 | 81.1 | 82.1 | |
| FlashRL (TIS)Backbone=LLaDA-8B, Evaluation Protocol=Full-model fine-tuning2026.05 | 74.37 | 78.24 | 80.44 | |
| FP8 RolloutBackbone=LLaDA-8B, Evaluation Protocol=Full-model fine-tuning2026.05 | 73.99 | 79.15 | 80.89 |