SpecRouter: Adaptive Routing for Multi-Level Speculative Decoding in Large Language Models

TL;DR

SpecRouter introduces an adaptive, multi-level speculative decoding framework for large language models, dynamically optimizing inference paths based on real-time feedback to balance quality and latency.

Contribution

It proposes a novel adaptive routing approach with multi-level verification and synchronized state management to improve LLM inference efficiency.

Findings

Reduces inference latency compared to static methods.

Effectively balances quality and speed through dynamic model chaining.

Demonstrates promising preliminary experimental results.

Abstract

Large Language Models (LLMs) present a critical trade-off between inference quality and computational cost: larger models offer superior capabilities but incur significant latency, while smaller models are faster but less powerful. Existing serving strategies often employ fixed model scales or static two-stage speculative decoding, failing to dynamically adapt to the varying complexities of user requests or fluctuations in system performance. This paper introduces \systemname{}, a novel framework that reimagines LLM inference as an adaptive routing problem solved through multi-level speculative decoding. \systemname{} dynamically constructs and optimizes inference "paths" (chains of models) based on real-time feedback, addressing the limitations of static approaches. Our contributions are threefold: (1) An \textbf{adaptive model chain scheduling} mechanism that leverages performance profiling (execution times) and predictive similarity metrics (derived from token distribution divergence) to continuously select the optimal sequence of draft and verifier models, minimizing predicted latency per generated token. (2) A \textbf{multi-level collaborative verification} framework where intermediate models within the selected chain can validate speculative tokens, reducing the verification burden on the final, most powerful target model. (3) A \textbf{synchronized state management} system providing efficient, consistent KV cache handling across heterogeneous models in the chain, including precise, low-overhead rollbacks tailored for asynchronous batch processing inherent in multi-level speculation. Preliminary experiments demonstrate the validity of our method.