SOURCE-LINKED INTELLIGENCE

GrowMTP: Can RL Grow Its Own Draft Head?

arXiv · AI, language, vision and robotics · article · Sep 15, 2026 · UTC

Reinforcement learning (RL) post-training drives the frontier capabilities of large language models, with its wall-clock dominated by autoregressive rollout generation. Speculative decoding is an established remedy for this bottleneck, but existing draft heads must be pretrained or warmed up before RL, introducing substantial training cost outside the RL run to be accelerated. We observe that RL training itself provides both conditions required for online draft-head training: its rollout distribution is far narrower than that of pretraining, and its verification step continuously produces supe

Read original source ↗ Open in workspace

recordType

paper

region

Global