1
When RL Meets Adaptive Speculative Training: A Unified Training-Serving System
强化学习驱动自适应投机训练，统一训练与推理流程，消除部署延迟，加速大模型服务。
arXiv:2602.06932v4 Announce Type: replace Abstract: Speculative decoding can significantly accelerate LLM serving, yet most deployments today disentan…