Learning to Draft: Adaptive Speculative Decoding with Reinforcement Learning
This work addresses the efficiency bottleneck in LLM inference for real-time applications, representing a novel method rather than an incremental improvement.
The paper tackles the problem of accelerating large language model inference by optimizing the trade-off between drafting and verifying tokens in speculative decoding, achieving speedup ratios of 2.24x to 4.32x and outperforming the state-of-the-art method by up to 36.4%.
Speculative decoding accelerates large language model (LLM) inference by using a small draft model to generate candidate tokens for a larger target model to verify. The efficacy of this technique hinges on the trade-off between the time spent on drafting candidates and verifying them. However, current state-of-the-art methods rely on a static time allocation, while recent dynamic approaches optimize for proxy metrics like acceptance length, often neglecting the true time cost and treating the drafting and verification phases in isolation. To address these limitations, we introduce Learning to Draft (LTD), a novel method that directly optimizes for throughput of each draft-and-verify cycle. We formulate the problem as a reinforcement learning environment and train two co-adaptive policies to dynamically coordinate the draft and verification phases. This encourages the policies to adapt to each other and explicitly maximize decoding efficiency. We conducted extensive evaluations on five diverse LLMs and four distinct tasks. Our results show that LTD achieves speedup ratios ranging from 2.24x to 4.32x, outperforming the state-of-the-art method Eagle3 up to 36.4%.