List your company

Mathematical Reasoning on MATH 500 (Token Length Accuracy)

33.8Accuracy (128 tokens)BF16
30.6831.4932.333.11May 13, 2026

Evaluation results

MethodLinks
BF16Backbone=LLaDA-8B, Evaluation Protocol=Full-model fine-tuning2026.0533.838.9140.2
FlashRL (TIS)Backbone=LLaDA-8B, Evaluation Protocol=Full-model fine-tuning2026.0533.6837.439
AISBackbone=LLaDA-8B, Evaluation Protocol=Full-model fine-tuning2026.0532.838.640.27
FP8 RolloutBackbone=LLaDA-8B, Evaluation Protocol=Full-model fine-tuning2026.0530.83737.8