用于推测解码的验证感知训练

摘要

投机式解码通过使用一个草稿模型生成候选词元，再由目标模型在单次前向传播中加以验证，从而加速大语言模型推理。验证过程按顺序进行，并且从首个被拒绝的位置起丢弃其后的所有位置；然而，现有的草稿训练依赖于对目标的词元级模仿，并采用固定的逐位置权重，既不反映上述任一性质。我们提出验证感知训练（VAT），一种即插即用框架，它在每个训练步骤模拟验证过程，并将由此产生的接受与拒绝模式转化为监督信号。VAT由两个组件构成：（i）验证头，一个轻量的联合训练二分类器，用于监督草稿模型判断每个位置能否通过顺序验证；（ii）验证自适应加权，它取代固定的加权方案，在样本首个拒绝点之前保持全权重，并将衰减的起始位置重新锚定到该拒绝点。VAT仅修改训练目标，因此可以叠加在现有方法之上，而无需改变草稿架构、目标模型或推理流程。将VAT应用于EAGLE-3和DFlash，在Qwen3-4B、Qwen3-8B和LLaMA-3.1-8B上，平均接受长度最多提升11.4%，端到端加速最多提升8.7%，并在数学、代码和聊天基准上均获得一致改进。代码将发布在 https://github.com/naver-ai/vat。

English

Speculative decoding accelerates large language model inference by using a draft model to generate candidate tokens, which are verified by the target model in a single forward pass. Verification proceeds sequentially and discards every position from the first rejection onward, yet existing draft training relies on token-level imitation of the target with a fixed per-position weighting that reflects neither property. We introduce Verification-Aware Training (VAT), a plug-in framework that simulates verification at every training step and turns the resulting accept and reject patterns into supervision. VAT consists of two components: (i) a verification head, a lightweight jointly trained binary classifier that supervises the draft model on whether each position survives sequential verification; (ii) verification-adaptive weighting, which replaces the fixed weighting schedule by keeping full weight up to each sample's first rejection point and re-anchoring the decay to start there. VAT modifies only the training objective, so it can be layered on top of existing methods without changing the draft architecture, the target model, or the inference procedure. Applied to EAGLE-3 and DFlash on Qwen3-4B, Qwen3-8B, and LLaMA-3.1-8B, VAT improves average acceptance length by up to 11.4% and wall-clock speedup by up to 8.7%, with consistent gains across math, code, and chat benchmarks. Code will be available at https://github.com/naver-ai/vat