Speculative Decoding Speed-of-Light: Optimal Lower Bounds via Branching Random Walks
This work addresses a foundational problem for researchers and practitioners in AI by providing theoretical insights into the limits of parallel token generation, though it is incremental as it builds on existing speculative decoding techniques.
The paper tackled the problem of understanding the fundamental limits on speedup achievable by speculative decoding in large language models, establishing the first tight lower bounds on runtime for deterministic algorithms and validating these bounds empirically on Llama models.
Speculative generation has emerged as a promising technique to accelerate inference in large language models (LLMs) by leveraging parallelism to verify multiple draft tokens simultaneously. However, the fundamental limits on the achievable speedup remain poorly understood. In this work, we establish the first ``tight'' lower bounds on the runtime of any deterministic speculative generation algorithm. This is achieved by drawing a parallel between the token generation process and branching random walks, which allows us to analyze the optimal draft tree selection problem. We prove, under basic assumptions, that the expected number of tokens successfully predicted per speculative iteration is bounded as $\mathbb{E}[X] \leq (μ+ μ_{(2)})\log(P )/μ^2 + O(1)$, where $P$ is the verifier's capacity, $μ$ is the expected entropy of the verifier's output distribution, and $μ_{(2)}$ is the expected second log-moment. This result provides new insights into the limits of parallel token generation, and could guide the design of future speculative decoding systems. Empirical evaluations on Llama models validate our theoretical predictions, confirming the tightness of our bounds in practical settings.