AdaSwitch: Adaptive Switching Generation for Knowledge Distillation

• URL: http://arxiv.org/abs/2510.07842v1
• Date: Thu, 09 Oct 2025 06:38:37 GMT
• Title: AdaSwitch: Adaptive Switching Generation for Knowledge Distillation
• Authors: Jingyu Peng, Maolin Wang, Hengyi Cai, Yuchen Li, Kai Zhang, Shuaiqiang Wang, Dawei Yin, Xiangyu Zhao,
• Abstract summary: Small language models (SLMs) are crucial for applications with strict latency and computational constraints.<n>We propose AdaSwitch, a novel approach that combines on-policy and off-policy generation at the token level.<n>AdaSwitch consistently improves accuracy, offering a practical and effective method for distilling SLMs with acceptable additional overhead.
• Score: 58.647880811071495
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Small language models (SLMs) are crucial for applications with strict latency and computational constraints, yet achieving high performance remains challenging. Knowledge distillation (KD) can transfer capabilities from large teacher models, but existing methods involve trade-offs: off-policy distillation provides high-quality supervision but introduces a training-inference mismatch, while on-policy approaches maintain consistency but rely on low-quality student outputs. To address these issues, we propose AdaSwitch, a novel approach that dynamically combines on-policy and off-policy generation at the token level. AdaSwitch allows the student to first explore its own predictions and then selectively integrate teacher guidance based on real-time quality assessment. This approach simultaneously preserves consistency and maintains supervision quality. Experiments on three datasets with two teacher-student LLM pairs demonstrate that AdaSwitch consistently improves accuracy, offering a practical and effective method for distilling SLMs with acceptable additional overhead.

Related papers

• From Reasoning LLMs to BERT: A Two-Stage Distillation Framework for Search Relevance [20.096802351171377]
  e-commerce search systems face strict latency requirements that prevent the direct application of Large Language Models.<n>We propose a two-stage reasoning distillation framework to transfer reasoning capabilities from a powerful teacher LLM to a lightweight, deployment-friendly student model.<n>Our framework achieves significant improvements across multiple metrics, validating its effectiveness and practical value.
  arXiv  Detail & Related papers  (2025-10-13T06:46:43Z)
• CKAA: Cross-subspace Knowledge Alignment and Aggregation for Robust Continual Learning [80.18781219542016]
  Continual Learning (CL) empowers AI models to continuously learn from sequential task streams.<n>Recent parameter-efficient fine-tuning (PEFT)-based CL methods have garnered increasing attention due to their superior performance.<n>We propose Cross-subspace Knowledge Alignment and Aggregation (CKAA) to enhance robustness against misleading task-ids.
  arXiv  Detail & Related papers  (2025-07-13T03:11:35Z)
• Teach2Eval: An Indirect Evaluation Method for LLM by Judging How It Teaches [46.0474342507327]
  We introduce Teach2Eval, an indirect evaluation framework inspired by the Feynman Technique.<n>Our method evaluates a model's multiple abilities to teach weaker student models to perform tasks effectively.
  arXiv  Detail & Related papers  (2025-05-18T06:51:10Z)
• JointDistill: Adaptive Multi-Task Distillation for Joint Depth Estimation and Scene Segmentation [31.89422375115854]
  This work explores how the multi-task distillation could be used to improve unified modeling.<n>We propose a self-adaptive distillation method that can adjust the knowledge amount from each teacher according to the student's current learning ability.<n>We evaluate our method on multiple benchmarking datasets including Cityscapes and NYU-v2.
  arXiv  Detail & Related papers  (2025-05-15T08:00:48Z)
• Learning from Stochastic Teacher Representations Using Student-Guided Knowledge Distillation [64.15918654558816]
  Self-distillation (SSD) training strategy is introduced for filtering and weighting teacher representation to distill from task-relevant representations only.<n> Experimental results on real-world affective computing, wearable/biosignal datasets from the UCR Archive, the HAR dataset, and image classification datasets show that the proposed SSD method can outperform state-of-the-art methods.
  arXiv  Detail & Related papers  (2025-04-19T14:08:56Z)
• Cross-View Consistency Regularisation for Knowledge Distillation [13.918476599394603]
  This work is inspired by the success of cross-view learning in fields such as semi-supervised learning.<n>We introduce within-view and cross-view regularisations to standard logit-based distillation frameworks.<n>We also perform confidence-based soft label mining to improve the quality of distilling signals from the teacher.
  arXiv  Detail & Related papers  (2024-12-21T05:41:47Z)
• CoDTS: Enhancing Sparsely Supervised Collaborative Perception with a Dual Teacher-Student Framework [15.538850922083652]
  We propose an end-to-end Collaborative perception Dual Teacher-Student framework (CoDTS)<n>It employs adaptive complementary learning to produce both high-quality and high-quantity pseudo labels.<n>CoDTS effectively ensures an optimal balance of pseudo labels in both quality and quantity.
  arXiv  Detail & Related papers  (2024-12-11T12:34:37Z)
• Distillation Matters: Empowering Sequential Recommenders to Match the Performance of Large Language Model [12.6937643116018]
  Large Language Models (LLMs) have been effectively utilized as recommenders, achieving impressive performance. However, the high inference latency of LLMs significantly restricts their practical deployment. This work investigates knowledge distillation from cumbersome LLM-based recommendation models to lightweight sequential models.
  arXiv  Detail & Related papers  (2024-05-01T06:23:54Z)
• Distantly-Supervised Named Entity Recognition with Adaptive Teacher Learning and Fine-grained Student Ensemble [56.705249154629264]
  Self-training teacher-student frameworks are proposed to improve the robustness of NER models. In this paper, we propose an adaptive teacher learning comprised of two teacher-student networks. Fine-grained student ensemble updates each fragment of the teacher model with a temporal moving average of the corresponding fragment of the student, which enhances consistent predictions on each model fragment against noise.
  arXiv  Detail & Related papers  (2022-12-13T12:14:09Z)
• Parameter-Efficient and Student-Friendly Knowledge Distillation [83.56365548607863]
  We present a parameter-efficient and student-friendly knowledge distillation method, namely PESF-KD, to achieve efficient and sufficient knowledge transfer. Experiments on a variety of benchmarks show that PESF-KD can significantly reduce the training cost while obtaining competitive results compared to advanced online distillation methods.
  arXiv  Detail & Related papers  (2022-05-28T16:11:49Z)
• Transfer Heterogeneous Knowledge Among Peer-to-Peer Teammates: A Model Distillation Approach [55.83558520598304]
  We propose a brand new solution to reuse experiences and transfer value functions among multiple students via model distillation. We also describe how to design an efficient communication protocol to exploit heterogeneous knowledge. Our proposed framework, namely Learning and Teaching Categorical Reinforcement, shows promising performance on stabilizing and accelerating learning progress.
  arXiv  Detail & Related papers  (2020-02-06T11:31:04Z)

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.