DeLS-Spec: Decoupled Long-Short Contexts for Parallel Speculative Drafting

Researchers propose DeLS-Spec, a speculative decoding method that decouples long- and short-context experts to accelerate LLM inference. By adding a lightweight local head to the existing DFlash model, DeLS-Spec achieves higher speedup and acceptance length on Qwen3 models without costly joint training. The method offers modular flexibility and low training cost.

arXiv:2607.07409v1 Announce Type: new Abstract: Speculative decoding accelerates LLM inference by drafting multiple tokens and verifying them in parallel. Block-parallel drafters such as DFlash further improve drafting efficiency by predicting an entire block in one pass, but their position-wise predictions lack explicit intra-block causal conditioning. Recent methods such as Domino and DSpark attempt to introduce such causality into block-parallel drafting, but they require training the draft model from scratch, which limits their flexibility and increases training cost. We propose DeLS-Spec, a decoupled long-short context speculative decoding method. DeLS-Spec treats the fixed DFlash model as a long-context expert and introduces a lightweight local head as a short-context expert. The local head can be trained independently with a standard next-token prediction objective, without joint training with the target model or the DFlash backbone, leading to extremely low training cost. At inference time, DeLS-Spec combines long-context and short-context logits, and the local head is not tied to a specific DFlash checkpoint, making the method more modular and flexible. Experiments on Qwen3 models show that DeLS-Spec consistently improves speedup and average acceptance length over DFlash across math, code, and dialogue benchmarks.