SpecKD: Speculative Decoding for Effective Knowledge Distillation of LLMs

• URL: http://arxiv.org/abs/2510.24021v1
• Date: Tue, 28 Oct 2025 03:02:22 GMT
• Title: SpecKD: Speculative Decoding for Effective Knowledge Distillation of LLMs
• Authors: Haiduo Huang, Jiangcheng Song, Yadong Zhang, Pengju Ren,
• Abstract summary: Speculative Knowledge Distillation (SpecKD) is a novel, plug-and-play framework that introduces a dynamic, token-level gating mechanism.<n>SpecKD consistently and significantly outperforms strong Knowledge Distillation baselines.
• Score: 7.838632409628936
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Knowledge Distillation (KD) has become a cornerstone technique for compressing Large Language Models (LLMs) into smaller, more efficient student models. However, conventional KD approaches typically apply the distillation loss uniformly across all tokens, regardless of the teacher's confidence. This indiscriminate mimicry can introduce noise, as the student is forced to learn from the teacher's uncertain or high-entropy predictions, which may ultimately harm student performance-especially when the teacher is much larger and more powerful. To address this, we propose Speculative Knowledge Distillation (SpecKD), a novel, plug-and-play framework that introduces a dynamic, token-level gating mechanism inspired by the "propose-and-verify" paradigm of speculative decoding. At each step, the student's token proposal is verified against the teacher's distribution; the distillation loss is selectively applied only to "accepted" tokens, while "rejected" tokens are masked out. Extensive experiments on diverse text generation tasks show that SpecKD consistently and significantly outperforms strong KD baselines, leading to more stable training and more capable student models, and achieving state-of-the-art results.

Related papers

• Speculative Knowledge Distillation: Bridging the Teacher-Student Gap Through Interleaved Sampling [81.00825302340984]
  We introduce Speculative Knowledge Distillation (SKD) to generate high-quality training data on-the-fly.<n>In SKD, the student proposes tokens, and the teacher replaces poorly ranked ones based on its own distribution.<n>We evaluate SKD on various text generation tasks, including translation, summarization, math, and instruction following.
  arXiv  Detail & Related papers  (2024-10-15T06:51:25Z)
• Linear Projections of Teacher Embeddings for Few-Class Distillation [14.99228980898161]
  Knowledge Distillation (KD) has emerged as a promising approach for transferring knowledge from a larger, more complex teacher model to a smaller student model. We introduce a novel method for distilling knowledge from the teacher's model representations, which we term Learning Embedding Linear Projections (LELP) Our experimental evaluation on large-scale NLP benchmarks like Amazon Reviews and Sentiment140 demonstrate the LELP is consistently competitive with, and typically superior to, existing state-of-the-art distillation algorithms for binary and few-class problems.
  arXiv  Detail & Related papers  (2024-09-30T16:07:34Z)
• PromptKD: Distilling Student-Friendly Knowledge for Generative Language Models via Prompt Tuning [30.70974942397732]
  We propose PromptKD to enable generative language models to transfer student-friendly knowledge. Experiments on instruction-following datasets show that PromptKD achieves state-of-the-art performance. Further analysis suggests that distilling student-friendly knowledge alleviates exposure bias effectively throughout the entire training process.
  arXiv  Detail & Related papers  (2024-02-20T09:10:08Z)
• Revisiting Knowledge Distillation for Autoregressive Language Models [88.80146574509195]
  We propose a simple yet effective adaptive teaching approach (ATKD) to improve the knowledge distillation (KD) The core of ATKD is to reduce rote learning and make teaching more diverse and flexible. Experiments on 8 LM tasks show that, with the help of ATKD, various baseline KD methods can achieve consistent and significant performance gains.
  arXiv  Detail & Related papers  (2024-02-19T07:01:10Z)
• Comparative Knowledge Distillation [102.35425896967791]
  Traditional Knowledge Distillation (KD) assumes readily available access to teacher models for frequent inference. We propose Comparative Knowledge Distillation (CKD), which encourages student models to understand the nuanced differences in a teacher model's interpretations of samples. CKD consistently outperforms state of the art data augmentation and KD techniques.
  arXiv  Detail & Related papers  (2023-11-03T21:55:33Z)
• On-Policy Distillation of Language Models: Learning from Self-Generated Mistakes [44.97759066341107]
  Generalized Knowledge Distillation (GKD) trains the student on its self-generated output sequences by leveraging feedback from the teacher. We demonstrate the efficacy of GKD for distilling auto-regressive language models on summarization, translation, and arithmetic reasoning tasks.
  arXiv  Detail & Related papers  (2023-06-23T17:56:26Z)
• ALM-KD: Knowledge Distillation with noisy labels via adaptive loss mixing [25.49637460661711]
  Knowledge distillation is a technique where the outputs of a pretrained model are used for training a student model in a supervised setting. We tackle this problem via the use of an adaptive loss mixing scheme during KD. We demonstrate performance gains obtained using our approach in the standard KD setting as well as in multi-teacher and self-distillation settings.
  arXiv  Detail & Related papers  (2022-02-07T14:53:22Z)
• Learning to Teach with Student Feedback [67.41261090761834]
  Interactive Knowledge Distillation (IKD) allows the teacher to learn to teach from the feedback of the student. IKD trains the teacher model to generate specific soft target at each training step for a certain student. Joint optimization for both teacher and student is achieved by two iterative steps.
  arXiv  Detail & Related papers  (2021-09-10T03:01:01Z)
• Undistillable: Making A Nasty Teacher That CANNOT teach students [84.6111281091602]
  This paper introduces and investigates a concept called Nasty Teacher: a specially trained teacher network that yields nearly the same performance as a normal one. We propose a simple yet effective algorithm to build the nasty teacher, called self-undermining knowledge distillation.
  arXiv  Detail & Related papers  (2021-05-16T08:41:30Z)
• ProxylessKD: Direct Knowledge Distillation with Inherited Classifier for Face Recognition [84.49978494275382]
  Knowledge Distillation (KD) refers to transferring knowledge from a large model to a smaller one. In this work, we focus on its application in face recognition. We propose a novel method named ProxylessKD that directly optimize face recognition accuracy.
  arXiv  Detail & Related papers  (2020-10-31T13:14:34Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.