RL 训练加速是 LLM 后训练的核心痛点，Bebop 用 MTP+拒绝采样把加速做到 1.8 倍，做 RL 训练优化的团队可以直接参考其 TV 损失和离线训练策略。
强化学习（RL）训练中，rollout 阶段是主要瓶颈。多 Token 预测（MTP）本可通过推测解码加速，但 RL 训练中 MTP 接受率会因模型熵波动而显著下降。Bebop 研究揭示了熵与接受率的负线性关系，并提出概率拒绝采样可缓解熵干扰。他们进一步提出端到端 TV 损失函数，直接优化拒绝采样接受率，在数学推理、代码生成和智能体任务上实现最高 95% 接受率，吞吐量提升 25%。在 Qwen3.5/3.6/3.7 模型上，异步 RL 训练端到端加速达 1.8 倍，且无需在线更新 MTP。
Breaking Entropy Bounds: Accelerating RL Training via MTP with Rejection Sampling
Reinforcement learning (RL) has become a key component in modern large language models, yet the rollout stage remains the key bottleneck in RL training pipelines. Although Multi-Token Prediction (MTP) offers a natural solution to accelerate rollouts through speculative decoding, many studies have observed that MTP acceptance rates degrade significantly during RL training, leading to limited speedup performance. To address this bottleneck, we present Bebop, a systematic study of MTP in LLM post-training, and offer practical recipes to integrate MTP into large-scale RL pipelines. First, we reveal that the MTP acceptance rate is fundamentally bounded by the fluctuation of model entropy, which demonstrates a clear negative linear relationship with the rise of entropy in the RL stage. Second, we show that probabilistic rejection sampling largely alleviates the disturbance introduced by entropy in RL compared to greedy draft sampling. We further identify that the conventional MTP training objectives (cross-entropy or KL) are suboptimal in such settings, and therefore we propose a novel end-to-end TV loss that directly optimizes multi-step rejection sampling acceptance rate, yielding ~10% acceptance rate improvements, achieving up to 95% acceptance rates and up to 25% extra inference throughput gains across mathematical reasoning, code generation, and agentic tasks. Third, we test various online MTP training strategies during RL and show that pre-RL MTP training with e2e TV loss and rejection sampling achieves a consistent acceptance rate and speedup throughout the entire RL, eliminating the need for costly online MTP updating. We provide extensive experiments and analysis that validate our findings. Experimental results show our method achieves up to 1.8x end-to-end acceleration in async RL training of Qwen3.5, Qwen3.6, and Qwen3.7 models.