TRIAGE: Direction-Aware Mismatch Stabilization of Native NVFP4 Reinforcement Learning
Open paper
Benchmark
Model
Rank
Results
mathematical-reasoning-on-math-500
Qwen3-30B-A3B-Base (NVFP4+TRIAGE, step 1700)
#52
Accuracy: 91.6
Rank counts only results with a code link.