Knowledge Distillation

Ask about this topic

Overview

Knowledge distillation studies how a student model can learn from a teacher’s outputs, internal representations, examples, or behavior, often to reduce inference cost or transfer capabilities. Common approaches match softened class probabilities, hidden features, attention or relational structure, and use teacher-generated data; newer language-model methods also train on student-generated rollouts. The field spans model compression, cross-architecture and multimodal transfer, and settings such as federated learning, robotics, and low-resource tasks. A central challenge is choosing supervision that fits the student and remains reliable: stronger or denser teacher signals can help, but can also cause negative transfer, forgetting, or loss of capabilities not captured by the benchmark.

Latest papers

CardsList

829 papers164 papers in the last four weeks, up 95% on the four weeks before. 0.9% of all new papers.

May 8, 2026

Teacher-Feature Drifting: One-Step Diffusion Distillation with Pretrained Diffusion Representations

Sampling from pretrained diffusion and flow-matching models typically requires many forward passes to generate diverse and high-fidelity images. Existing distillation methods often rely on multiple auxiliary networks, carefully designed training stages, or complex optimization pipelines. In this work, we revisit the recently proposed Drifting Model objective and show that a single drifting loss can be directly used to simplify one step distillation. A key observation is that the pretrained diffusion teacher itself already provides a strong representation space. Unlike the original Drifting Model, which relies on an additional pretrained feature extractor, we use intermediate hidden states of the pretrained teacher model as the feature representation. This removes the need for training or introducing an extra representation network while preserving a semantically meaningful feature geometry for drifting. Furthermore, we introduce a lightweight mode coverage loss to mitigate mode collapse during distillation and encourage the student generator to cover diverse teacher-supported regions. Extensive experiments on ImageNet and SDXL demonstrate that our method achieves efficient one step generation with competitive image quality and diversity, achieving FID scores of 1.58 on ImageNet-64×\times64 and 18.4 on SDXL, while substantially simplifying the overall distillation framework.

Yuan Zhang, Chenyi Li, Guoqing Ma +7

JD Explore Academy, China · The Hong Kong University of Science and Technology · Tsinghua University +1

Diffusion Model DistillationDiffusion SamplingDiffusion-Based Image GenerationFeature-Based Knowledge DistillationKnowledge Distillation
AskPDF
May 8, 2026

Attention Transfer Is Not Universally Effective for Vision Transformers

A recent work shows that Attention Transfer, which transfers only the attention patterns from a pre-trained teacher Vision Transformer (ViT) to a randomly initialized standard student ViT, is sufficient to recover the full benefit of the teacher's pre-trained weights. We revisit this finding on a comprehensive benchmark of 20 teachers from 11 well-known ViT families and reveal that Attention Transfer is not universally effective. While 7 families transfer successfully, 4 consistently fail, falling up to 5.1% below the from-scratch no-transfer baseline. Further results demonstrate that this failure is family-consistent across model sizes, and persists under extended training durations, different transfer datasets, and out-of-distribution evaluations. Controlled analyses then consistently localize the problem to the attention-routing channel, indicating that the key issue is not whether the student can match the teacher's attention patterns, but whether the matched patterns remain functional for the student. Crucially, we identify architectural mismatch between the pre-trained teacher and the standard student as the primary mechanism. By adding only the teacher's native architectural components to the student in a randomly initialized state, we completely reverse the failure for all 4 families. Notably, these components alone do not improve from-scratch training, confirming that they specifically unlock the usability of the teacher's attention. We further show that this failure is not explained by the inadequate choice of transfer loss, differences in pre-training recipes, or the intrinsic quality of attention maps. Our findings refine the prevailing understanding of attention in ViT representations: attention is sufficient only when the student architecture matches the teacher.

Huaiyuan Qin, Muli Yang, Gabriel James Goenawan +4

Institute of Advanced Intelligence and Computing (IAIC), A*STAR, Singapore · School of Artificial Intelligence, Sichuan University · School of Automation and Intelligent Sensing, Shanghai Jiao Tong University

Self-AttentionVision TransformerTransfer LearningAttention MechanismsTeacher-Student LearningKnowledge Distillation
AskPDF
May 8, 2026

Structural Rationale Distillation via Reasoning Space Compression

When distilling reasoning from large language models (LLMs) into smaller ones, teacher rationales for similar problems often vary wildly in structure and strategy. Like a chef who makes the same dish differently each time, this inconsistency burdens the student with noisy supervision that is hard to internalize. We propose Distillation through Reasoning Path Compression (D-RPC), which constrains the teacher to follow a compact, dynamically maintained bank of reusable high-level reasoning paths. For each training question, D-RPC retrieves the most relevant path and conditions the teacher to follow it, producing rationales that are consistent across similar problems yet diverse enough to cover different problem types. A PAC-Bayes analysis formalizes the resulting trade-off between bank size and coverage: smaller banks reduce supervision entropy but risk coverage gaps, and the generalization bound identifies an optimal intermediate size confirmed by our ablations. Across five math and commonsense reasoning benchmarks with two student models, D-RPC consistently outperforms chain-of-thought distillation, freeform rationale generation, direct distillation, and structured-supervision baselines, while using fewer tokens than template-heavy alternatives.

Jialin Yang, Jiankun Wang, Jiajun Wu +3

University of Calgary · Calgary, Canada · University of Michigan +1

CoT DistillationStructured ReasoningTeacher-Student LearningKnowledge DistillationCoT CompressionLLM Reasoning
AskPDF
May 7, 2026

How to Compress KV Cache in RL Post-Training? Shadow Mask Distillation for Memory-Efficient Alignment

Reinforcement Learning (RL) has emerged as a crucial paradigm for unlocking the advanced reasoning capabilities of Large Language Models (LLMs), encompassing frameworks like RLHF and RLAIF. Regardless of the specific optimization algorithm (e.g., PPO, GRPO, or Online DPO), online RL inherently requires an exploratory trajectory generation (rollout) phase. However, for long-context reasoning tasks, this rollout phase imposes a severe ``memory wall'' due to the exorbitant Key-Value (KV) cache footprint. While applying KV cache compression during rollouts mitigates this memory overhead, it induces a critical off-policy bias. Although modern KV compression is often nearly lossless during standard inference, even minuscule approximation errors are drastically amplified by the inherent instability of RL optimization. Specifically, the sampler generates responses under a sparse context, whereas the learner updates parameters using the full, dense context. Existing statistical solutions, such as importance reweighting, struggle to correct this magnified bias, suffering from high gradient variance and severe sample inefficiency.

Rui Zhu, Weiheng Bai, Qiushi Wu +3

Yale University · University of Minnesota Twin Cities · Indiana University Bloomington

LLM AlignmentKV-Cache CompressionRL Post-TrainingKnowledge DistillationOff-Policy RL
AskPDF
May 7, 2026

Model-to-Data Distillation for Graph Neural Networks

Graph neural networks (GNNs) increasingly rely on sophisticated architectures and training procedures to achieve desirable properties such as high predictive performance, fairness, and robustness. However, these properties typically remain tied to the models that learn them, limiting their transferability to simpler models and downstream settings. We introduce model-to-data (M2D) distillation, a new distillation paradigm that transfers properties learned by a complex GNN teacher into graph data, enabling simpler models to recover them through standard training. M2D distillation explicitly trades model complexity for data complexity by jointly learning augmented node features and graph structure that encode the teacher's behavior. The resulting graph serves as a persistent medium for knowledge transfer and can be used with different downstream models. We show that M2D distillation enables simple GNNs to approximate the behavior of substantially more sophisticated teachers, including fairness-aware GNNs, Graph Attention Networks, and Graph Transformers, while maintaining comparable predictive performance and transferring desirable properties of the teacher.

Debolina Halder Lina, Arlei Silva

Department of Computer Science, Rice University, Houston, TX 77005, USA · Department of Computer Science & Ken Kennedy Institute, Rice University, Houston, TX 77005, USA

Graph Structure LearningGraph Neural NetworksDataset DistillationKnowledge Distillation
AskPDF
May 7, 2026

UniSD: Towards a Unified Self-Distillation Framework for Large Language Models

Self-distillation (SD) offers a promising path for adapting large language models (LLMs) without relying on stronger external teachers. However, SD in autoregressive LLMs remains challenging because self-generated trajectories are free-form, correctness is task-dependent, and plausible rationales can still provide unstable or unreliable supervision. Existing methods mainly examine isolated design choices, leaving their effectiveness, roles, and interactions unclear. In this paper, we propose UniSD, a unified framework to systematically study self-distillation. UniSD integrates complementary mechanisms that address supervision reliability, representation alignment, and training stability, including multi-teacher agreement, EMA teacher stabilization, token-level contrastive learning, feature matching, and divergence clipping. Across six benchmarks and six models from three model families, UniSD reveals when self-distillation improves over static imitation, which components drive the gains, and how these components interact across tasks. Guided by these insights, we construct UniSDfull, an integrated pipeline that combines complementary components and achieves the strongest overall performance, improving over the base model by +5.4 points and the strongest baseline by +2.8 points. Extensive evaluation highlights self-distillation as a practical and steerable approach for efficient LLM adaptation without stronger external teachers.

Yiqiao Jin, Yiyang Wang, Lucheng Fu +7

Georgia Institute of Technology · University of California, Los Angeles · Carnegie Mellon University +1

Language Model DistillationSelf-DistillationKnowledge Distillation
AskPDF
May 7, 2026

CKT-WAM: Parameter-Efficient Context Knowledge Transfer Between World Action Models

World action models (WAMs) provide a powerful generative framework for embodied control, yet transferring knowledge across heterogeneous WAMs remains challenging due to mismatched latent interfaces, high adaptation cost, and the rigidity of conventional distillation objectives. We propose CKT-WAM, a parameter-efficient Context Knowledge Transfer framework that transfers teacher WAM's knowledge into a student WAM through a compact context in the text embedding space, rather than output imitation or dense hidden-state matching. Specifically, CKT-WAM extracts intermediate teacher hidden states, reduces the number of tokens via compressors' learnable-query cross attention (LQCA), and transforms them through an always-on generalized adapter, a lightweight router, and sparsely activated specialized adapters. The resulting context is then appended to the student's conditioning textual embeddings, thereby injecting the transferred knowledge into the student with minimal architectural modification. Experiments show that CKT-WAM consistently improves zero-shot generalization and achieves the best overall performance on LIBERO-Plus, reaching 86.1% total success rate with only 1.17% trainable parameters, while approaching full fine-tuning performance. Beyond simulation, CKT-WAM also demonstrates strong real-world long-horizon manipulation ability, achieving the best average success rate of 83.3% across four multi-step and long-horizon tasks. Code is available at https://github.com/YuhuaJiang2002/CKT-WAM.

Yuhua Jiang, Yijun Guo, Hongbing Yang +7

Tsinghua University · LivsynRobotics · Shanghai AI Laboratory

AI ControlRobotic ManipulationWorld Action ModelsKnowledge DistillationEmbodied World ModelsParameter-Efficient Fine-Tuning
AskPDF
May 7, 2026

TinySSL: Distilled Self-Supervised Pretraining for Sub-Megabyte MCU Models

Self-supervised learning (SSL) has transformed representation learning for large models, yet remains unexplored for microcontroller (MCU)-class models with fewer than 500K parameters. We identify three obstacles at this scale -- projection head dominance, representation bottleneck, and augmentation sensitivity -- and propose Capacity-Aware Distilled Self-Supervised Learning (CA-DSSL), a teacher-guided framework that overcomes them without labels or text supervision. CA-DSSL combines asymmetric distillation from a frozen DINO ViT-S/16 teacher, multi-scale feature distillation for spatial representations, and a progressive augmentation curriculum. On a MobileNetV2-0.35 backbone (396K parameters) pretrained on CIFAR-100, CA-DSSL reaches 62.7 0.5% linear-probe accuracy (3-seed mean) -- surpassing SimCLR-Tiny by 18 pp, matching SEED (61.7%) with 10 fewer projection parameters (426K vs. 3.15M), and reaching 94.0% of a supervised upper bound. Standard SSL methods (BYOL-Tiny, DINO-Tiny) collapse entirely at this scale. On Pascal VOC detection, CA-DSSL achieves 2.3 the mAP of random initialization and +3 pp over SEED, though SimCLR-Tiny matches CA-DSSL on detection mAP. The deployed backbone occupies 378 KB (INT8) with no inference overhead from pretraining. Preliminary ImageNet-100 experiments reveal that CA-DSSL's advantage is specific to small-data regimes; scaling to ImageNet-1K is discussed as future work.

Bibin Wilson

Self-Supervised LearningSelf-Supervised Pre-TrainingTiny MLKnowledge DistillationSelf-Supervised Visual Representation Learning
AskPDF
May 7, 2026

Estimating the Black-box LLM Uncertainty with Distribution-Aligned Adversarial Distillation

Large language models (LLMs) have progressed rapidly in complex reasoning and question answering, yet LLM hallucination remains a central bottleneck that hinders practical deployment, especially for commercial black-box LLMs accessible only via APIs. Existing uncertainty quantification methods typically depend on computationally expensive multiple sampling or internal parameters, which prevents real-time estimation and fails to capture information implicit in the black-box reasoning process. To address this issue, we propose Distribution-Aligned Adversarial Distillation (DisAAD), which introduces a generation-discrimination architecture to guide a lightweight proxy model to learn the high-quality regions of the output distribution of the black-box LLM, thus effectively endowing it with the ability to know whether the black-box LLM knows or not. Subsequently, we use the proxy model to reproduce the specific responses of the black-box LLM and estimate the corresponding uncertainty based on evidence learning. Extensive experiments have verified the effectiveness and promise of our proposed method, indicating that a proxy model even one that only accounts for 1% of the target LLM's size can achieve reliable uncertainty quantification.

Huizi Cui, Huan Ma, Qilin Wang +2

School of Computer Science and Technology, Tianjin University, China · School of Future Technology, Tianjin University, China · Georgia Tech Shenzhen Institute, Tianjin University, China +1

Language Model DistillationLLM Uncertainty EstimationKnowledge DistillationAdversarial Knowledge Distillation
AskPDF
May 7, 2026

FedeKD: Energy-Based Gating for Robust Federated Knowledge Distillation under Heterogeneous Settings

Federated learning (FL) operates in heterogeneous environments, where variations in data distributions and asymmetric model design often result in negative transfer. While federated knowledge distillation (FKD) avoids direct model parameter sharing, existing methods typically rely on public datasets or assume that transferred knowledge is uniformly reliable, which limits their robustness in practice. This paper presents FedeKD, a reliability-aware FKD framework that makes sample-wise trust estimation an explicit component of knowledge transfer, without relying on additional public data. Each client maintains a high-capacity private model for local learning and a lightweight shared proxy model for cross-client knowledge exchange. During training, proxy models are aggregated on the server to form a global proxy, which is then used to guide updates of the private models. At the core of FedeKD is an energy-based gating mechanism that converts task-specific private-proxy disagreement into sample-wise trust weights for backward distillation. This mechanism enables sample-wise weighting of knowledge transfer, where the proxy model contributes more to reliable samples while down-weighting unreliable ones. Extensive experiments on six real-world datasets demonstrate that FedeKD significantly reduces negative transfer under heterogeneous settings while maintaining strong predictive performance.

Quang-Huy Nguyen, Jiaqi Wang, Wei-shinn Ku

Department of Computer Science and Software Engineering Auburn University Auburn, AL 36849

Federated Knowledge DistillationKnowledge DistillationNegative TransferFederated LearningHeterogeneous Federated Learning
AskPDF
May 6, 2026

Learned Neighbor Trust for Collaborative Deployment in Model-Agnostic Decentralized Learning

Many decentralized distillation methods are designed around training-time coordination, yet deploy each node in isolation even when more capable neighbors remain available at inference time. This is an incomplete objective for settings such as IoT, where devices are heterogeneous, data is scarce and skewed, and a node's strongest neighbors may far exceed its own local capacity. We study how nodes should train so that their predictions compose well at deployment, and how each node should learn whom to trust. Under a server-free, model-agnostic protocol where nodes exchange only queries and soft predictions, we propose Learned Neighbor Trust (LNTrust) wherein each node learns a compact trust function over its neighborhood from local validation evidence. This trust function gates auxiliary distillation during training and defines a deployment ensemble at inference, so that collaboration learned during training transfers directly to deployment. Across datasets and topologies, LNTrust improves deployed accuracy over the strongest output-only baseline by large margins while using significantly less communication than previous methods.

Michael Lanier, Luise Ge, Sastry Kompella +1

Department of Computer Science Washington University in St. Louis and Nexcepta

Ensemble LearningDecentralized LearningKnowledge Distillation
AskPDF
May 6, 2026

Deep Reprogramming Distillation for Medical Foundation Models

Medical foundation models pre-trained on large-scale datasets have shown powerful versatile performance. However, when adapting medical foundation models for specific medical scenarios, it remains the inevitable challenge due to the gap induced by the discrepancy between pre-training and downstream tasks, the real-world computation, and speed constraints. Relevant techniques that probably handle this challenge more or less suffer from some intrinsic limitations. For example, knowledge distillation (KD) assumes that teacher and student models share the same task, training strategy, and model structure family, while prevalent parameter-efficient fine-tuning (PEFT) fails to achieve personalized and lightweight deployment. Even the combination of PEFT and KD still struggles to resolve model structures and training strategies inconsistencies between teacher and student models, leading to inefficient knowledge transfer. In this study, we propose a novel framework called Deep Reprogramming Distillation (DRD) to combat the general adaptation challenge. Specifically, DRD introduces the novel reprogramming module that on the one side overcomes the domain and task discrepancy between pretraining and downstream scenarios, and on the other side builds the student-friendly efficient distillation from foundation models to lightweight downstream models. Furthermore, to mitigate variability under different training conditions, we design a centered kernel alignment (CKA) distillation method to promote robust knowledge transfer. Empirical results show that DRD surpasses previous PEFT and KD methods across 18 medical downstream tasks under different foundation models, covering various scenarios including 2D/3D classification and 2D/3D segmentation.

Siyuan Du, Yuhang Zhou, Haolin Li +5

School of Computer Science, Fudan University and with the Shanghai AI Laboratory, Shanghai, China · Shanghai AI Laboratory, Shanghai, China · Cooperative Medianet Innovation Center, Shanghai Jiao Tong University and with the Shanghai AI Laboratory, Shanghai, China +3

Medical Imaging Foundation ModelsVision Foundation Model AdaptationKnowledge Distillation
AskPDF
May 5, 2026

Real Image Denoising with Knowledge Distillation for High-Performance Mobile NPUs

While deep-learning-based image restoration has achieved unprecedented fidelity, deployment on mobile Neural Processing Units (NPUs) remains bottlenecked by operator incompatibility and memory-access overhead. We propose an NPU-aware hardware-algorithm co-design approach for real-world image denoising on mobile NPUs. Our approach employs a high-capacity teacher to supervise a lightweight student network specifically designed to leverage the tiled-memory architectures of modern mobile SoCs. By prioritizing NPU-native primitives -- standard 3x3 convolutions, ReLU activations, and nearest-neighbor upsampling -- and employing a progressive context expansion strategy (up to 1024x1024 crops), the model achieves 37.66 dB PSNR / 0.9278 SSIM on the validation benchmark and 37.58 dB PSNR / 0.9098 SSIM on the held-out test benchmark at full resolution (2432x3200) in the Mobile AI 2026 challenge. Following the official challenge rules, the inference runtime is measured under a standardized Full HD (1088x1920) protocol, where it runs in 34.0 ms on the MediaTek Dimensity 9500 and 46.1 ms on the Qualcomm Snapdragon 8 Elite NPU. We further reveal an "Inference Inversion" effect, where strict adherence to NPU-compatible operations enables dedicated NPU execution up to 3.88x faster than the integrated mobile GPU. The 1.96M-parameter student recovers 99.8% of the teacher's restoration quality via high-alpha knowledge distillation (alpha = 0.9), achieving a 21.2x parameter reduction while closing the PSNR gap from 1.63 dB to only 0.05 dB. These results establish hardware-aware distillation as an effective strategy for unifying high-fidelity denoising with practical deployment across diverse mobile NPU architectures. The proposed lightweight student model (LiteDenoiseNet) and its training statistics are provided in the NN Dataset, available at https://github.com/ABrain-One/NN-Dataset.

Faraz Kayani, Sarmad Kayani, Asad Ahmed +2

Computer Vision Lab, CAIDAS & IFI, University of Würzburg, Germany

Efficient Neural Network InferenceKnowledge DistillationOn-Device InferenceImage Denoising
AskPDF
May 5, 2026

A Unified Approach to Interpreting Knowledge Distillation for Large Language Models via Interactions

Despite the success of knowledge distillation (KD) in Large Language Models (LLMs), the underlying mechanism behind its efficacy remains unclear. In this paper, we propose a unified approach to explore the common mechanism of various KD methods using interactions. Specifically, we decompose the output score of the LLM into the sum of numerous interactions. Each interaction represents a nonlinear relationship involving a set of input variables (e.g., words). Based on the decomposed interactions, we discover that the common mechanism underlying various KD methods is the sparsification of interactions, i.e., student models retain fewer interactions for inference while suppressing other interactions to zero effects. Furthermore, we discover that the performance variance across different KD methods arises from their capabilities in handling complex interactions. A KD method typically yields better performance if it enables the student model to achieve higher sparsity of complex interactions. Motivated by these insights, we propose a plug-and-play loss function called Complex Interaction Penalty (CIP) to explicitly enforce the sparsity of complex interactions during the distillation process. Extensive experiments demonstrate that integrating CIP consistently improves the performance of diverse KD methods on both in-domain and out-of-distribution benchmarks.

Qingzhuo Wang, Ruiyang Qin, Zhenxin Qin +2

School of Computer Science and Technology, Tongji University, Shanghai, China.

Feature Interaction ModelingLLM InterpretabilityLanguage Model DistillationKnowledge Distillation
AskPDF
May 5, 2026

Dynamic Distillation and Gradient Consistency for Robust Long-Tailed Incremental Learning

The task of Long-tailed Class Incremental Learning (LT-CIL) addresses the sequential learning of new classes from datasets with imbalanced class distributions. This scenario intensifies the fundamental problem of catastrophic forgetting, inherent to continual learning, with the dual challenges of under-learning minority classes and overfitting majority classes. To tackle these combined issues, this paper proposes two main techniques. First, we introduce gradient consistency regularization, which leverages the moving average of gradients to suppress abrupt fluctuations and stabilize the training process. Second, we dynamically adjust the weight of the distillation loss by measuring the degree of class imbalance with normalized entropy. This adaptive weighting establishes an optimal balance between retaining old knowledge and acquiring new information. Experiments on the CIFAR-100-LT, ImageNetSubset-LT, and Food101-LT benchmarks show that our method achieves consistent accuracy improvements of up to 5.0%. Furthermore, we demonstrate dramatic gains in the challenging 'In-ordered' setting, where tasks progress from majority to minority classes, highlighting our method's robustness in mitigating forgetting under unfavorable learning dynamics. This enhanced performance is achieved without a significant increase in computational overhead, demonstrating the practicality of our framework.

Taigo Sakai, Kazuhiro Hotta

Meijo University, 1-501 Shiogamaguchi, Tempaku-ku, Nagoya 468-8502, Japan

Class-Incremental LearningLong-Tail LearningAdaptive Loss WeightingIncremental LearningKnowledge DistillationConsistency Regularization
AskPDF
May 4, 2026

Standing on the Shoulders of Giants: Stabilized Knowledge Distillation for Cross--Language Code Clone Detection

Cross-language code clone detection (X-CCD) is challenging because semantically equivalent programs written in different languages often share little surface similarity. Although large language models (LLMs) have shown promise for semantic clone detection, their use as black-box systems raises concerns about cost, reproducibility, privacy, and unreliable output formatting. In particular, compact open-source models often struggle to follow reasoning-oriented prompts and to produce outputs that can be consistently mapped to binary clone labels. To address these limitations, we propose a knowledge distillation framework that transfers reasoning capabilities from DeepSeek-R1 into compact open-source student models for X-CCD. Using cross-language code pairs derived from Project CodeNet, we construct reasoning-oriented synthetic training data and fine-tune Phi3 and Qwen-Coder with LoRA adapters. We further introduce response stabilization methods, including forced conclusion prompting, a binary classification head, and a contrastive classification head, and evaluate model behavior using both predictive metrics and response rate. Experiments on Python--Java, Rust--Java, Rust--Python, and Rust--Ruby show that knowledge distillation consistently improves the reliability of compact models and often improves predictive performance, especially under distribution shift. In addition, classification-head variants substantially reduce inference time compared to generation-based inference. Overall, our results show that reasoning-oriented distillation combined with response stabilization makes compact open-source models more practical and reliable for X-CCD detection.

Mohamad Khajezade, Fatemeh H. Fard, Mohamed Sami Shehata

University of British Columbia, Okanagan Campus, Kelowna, Canada · University of British Columbia, Okanagan Campus, Kelowna, British Columbia, Canada

Language Model DistillationKnowledge DistillationCode Language Models
AskPDF
May 4, 2026

Combining Trained Models in Reinforcement Learning

Deep reinforcement learning (DRL) has delivered strong results in domains such as Atari and Go, but it still suffers from high sample cost and weak transfer beyond the training setting. A common response is to reuse information from previously trained models through transfer, distillation, ensemble methods, or federated training instead of learning each target task from random initialization. The literature on these mechanisms is fragmented, and published comparisons are hard to interpret because tasks, baselines, and compute budgets differ. This paper presents a PRISMA-guided systematic review of empirical studies on pretrained knowledge reuse in DRL. Starting from 589 records retrieved from IEEE Xplore, the ACM Digital Library, and citation tracing, we screened 570 unique records and assessed 89 full texts. After applying the final eligibility criteria, 15 empirical studies remained in the main synthesis. We analyzed them qualitatively across three factors: source-target similarity, diversity among reused models, and the fairness of comparisons against from-scratch baselines. Three patterns recur across the surviving corpus. First, positive results are concentrated in settings where source and target tasks share substantial structure or where the method includes an explicit gating or alignment mechanism. Second, evidence for ensembles and federated aggregation is promising but sparse and mostly limited to narrow settings. Third, compute-matched comparisons are rare, which weakens claims about efficiency gains over stronger single-agent baselines. The paper contributes a narrower and internally consistent review scope, a study-level synthesis of empirical evidence, and a provisional independence spectrum that should be treated as a hypothesis for future benchmarking rather than a validated metric.

Ujjwal Patil, Javad Ghofrani

Department of Computer Science Hochschule Bonn-Rhein-Sieg (H-BRS) Sankt Augustin, Germany

Reinforcement LearningKnowledge Distillation
AskPDF
May 3, 2026

EGAD: Entropy-Guided Adaptive Distillation for Token-Level Knowledge Transfer

Large language models (LLMs) have achieved remarkable performance across diverse domains, yet their enormous computational and memory requirements hinder deployment in resource-constrained environments. Knowledge distillation offers a promising solution by transferring knowledge from a large teacher model to a smaller student model. However, existing distillation methods typically treat all tokens equally, ignoring the fact that different tokens contribute unequally to model decisions. This can lead to inefficient knowledge transfer and reduced learning effectiveness. To address this limitation, we propose an entropy-based adaptive distillation strategy that dynamically adjusts the training process at the token level. Our method leverages the teacher's output entropy to guide three aspects of distillation. Specifically, we introduce a token-level curriculum by dynamically shifting focus from low- to high-entropy tokens during training. We further adjust the distillation temperature based on token entropy to better capture teacher confidence patterns. Moreover, we employ a dual-branch architecture for efficient logits-only distillation on easy tokens and deeper feature-based distillation on difficult tokens. Extensive experiments validate the soundness and effectiveness of our method.

Hao Zhang, Zhibin Zhang, Guangxin Wu +3

School of Advanced Interdisciplinary Sciences, University of Chinese Academy of Sciences · State Key Laboratory of AI Safety, Institute of Computing Technology, Chinese Academy of Sciences · University of Chinese Academy of Sciences

LLM CompressionLanguage Model DistillationKnowledge Distillation
AskPDF
May 2, 2026

Multi-Dataset Cross-Domain Knowledge Distillation for Unified Medical Image Segmentation, Classification, and Detection

We propose a unified cross-domain transfer learning framework that leverages knowledge from multiple heterogeneous medical imaging datasets to improve performance across segmentation, classification, and object detection tasks. Our approach employs a teacher-student paradigm in which a joint teacher model aggregates domain-invariant representations learned from diverse source datasets, while a task-specific student model is trained via multi-level knowledge distillation. Originally developed for medical image segmentation, the framework is extended to support image-level classification and object-level detection, enabling a general multi-task formulation for medical image analysis. We evaluate our method on a broad suite of datasets, including six segmentation benchmarks, BrainMetShare, ISLES, BraTS (MRI) and Lung MSD, LiTS, KiTS (CT), as well as multiple classification datasets for pulmonary disease and dementia, and detection datasets with native bounding-box annotations. Across all tasks and modalities, the proposed approach yields consistent improvements over strong dataset-specific and multi-head baselines, demonstrating enhanced robustness to distributional shifts and superior generalization. These findings highlight the potential of multi-dataset knowledge distillation as a scalable and task-agnostic approach for enhancing segmentation, classification, and object detection performance across heterogeneous medical imaging domains.

Ceausescu Ciprian-Mihai, Anghelina Ion-Marian, Alexe Dumitru-Bogdan

Faculty of Mathematics and Computer Science, University of Bucharest, Str. Academiei, 14, Bucharest, 010014, Romania · Gheorghe Mihoc-Caius Iacob Institute of Mathematical Statistics and Applied Mathematics of the Romanian Academy, Romania, Calea 13 Septembrie, 13, Bucharest, 050711, Romania

Medical Image Object DetectionMulti-Task LearningMedical Image ClassificationCross-Domain Transfer LearningMedical Image SegmentationObject Detection+1
AskPDF
May 2, 2026

MTA: Multi-Granular Trajectory Alignment for Large Language Model Distillation

Knowledge distillation is a key technique for compressing large language models (LLMs), but most existing methods align representations at fixed layers or token-level outputs, ignoring how representations evolve across depth. As a result, the student is only weakly guided to capture the teacher's internal relational structure during distillation, which limits knowledge transfer. To address this limitation, we propose Multi-Granular Trajectory Alignment (MTA), a framework that aligns teacher and student representations along their layer-wise transformation trajectory. MTA adopts a layer-adaptive strategy: lower layers are aligned at the word level to preserve lexical information, while higher layers operate on phrase-level spans (e.g., noun and verb phrases) to capture compositional semantics. We instantiate this idea through a Dynamic Structural Alignment loss that matches the relative geometry among semantic units within each layer. This design is motivated by empirical findings that Transformer representations become increasingly abstract with depth, and is also consistent with linguistic views in which higher-level meaning emerges through the composition of lower-level lexical units. We further incorporate a Hidden Representation Alignment loss to directly align selected teacher-student layers. Experiments show that MTA consistently outperforms state-of-the-art baselines on standard benchmarks, with ablations confirming the contribution of each component.

Pham Khanh Chi, Quoc Phong Dao, Thuat Nguyen +3

1Hanoi University of Science and Technology · 2Monash University · University of Oregon

Representation LearningLanguage Model DistillationRelational Knowledge DistillationKnowledge Distillation
AskPDF
May 2, 2026

AgriKD: Cross-Architecture Knowledge Distillation for Efficient Leaf Disease Classification

Automated leaf disease classification is critical for early disease detection in resource-constrained field environments. Vision Transformers (ViTs) provide strong representation capability by modeling long-range dependencies and inter-class relationships; however, their high computational cost makes them impractical for deployment on edge devices. As a result, existing approaches struggle to effectively transfer these rich representations to lightweight models. This paper introduces AgriKD, a cross-architecture knowledge distillation framework for efficient edge deployment, which transfers knowledge from a Vision Transformer (ViT) teacher to a compact convolutional student model. To bridge the representational gap between Transformer and CNN architectures, the proposed approach integrates multiple distillation objectives at the output, feature, and relational levels, where each objective captures a different aspect of the teacher knowledge. This enables the student model to better preserve and utilize transformer-derived global representations. Experiments on multiple leaf disease datasets show that the distilled student achieves performance comparable to the teacher while significantly improving efficiency, reducing model parameters by approximately 172 times, computational cost by 47.57 times, and inference latency by 18-22 times. Furthermore, the optimized model is deployed across multiple runtime formats, including ONNX, TFLite Float16, and TensorRT FP16, achieving consistent predictive performance with negligible accuracy degradation. Real-world deployment on NVIDIA Jetson edge devices and a mobile application demonstrates reliable real-time inference, highlighting the practicality of AgriKD for AI-powered agricultural applications in resource-constrained environments.

Minh-Dung Le, Minh-Duc Hoang, Hoang-Vu Truong +1

AIT laboratory, Faculty of Artificial Intelligence, FPT University, Da Nang, 550000, Viet Nam

Efficient Neural Network InferencePlant Disease ClassificationCross-Architecture Knowledge DistillationKnowledge DistillationOn-Device Inference
AskPDF
May 2, 2026

SRA: Span Representation Alignment for Large Language Model Distillation

Cross-Tokenizer Knowledge Distillation (CTKD) enables knowledge transfer between a large language model and a smaller student, even when they employ different tokenizers. While existing approaches mainly focus on token-level alignment strategies, which are often brittle and sensitive to discrepancies between tokenizers, we argue that the method of aggregating tokens into more robust representations before distillation is of equal importance. In this paper, we introduce SRA (Span Representation Alignment for Large Language Model Distillation), a novel framework that reframes CTKD through the physical lens of Multi-Particle Dynamical Systems. SRA shifts the fundamental unit of alignment from tokens to robust, tokenizer-agnostic spans. We model each span as a cluster of particles and represent its state by its Center of Mass (CoM) - an attention-weighted average that captures rich semantic information. We leverage the concept of span centers of mass with attention-derived weighting to prioritize the most salient spans. In addition, we employ a geometric regularizer to preserve the structural integrity of the representation space and introduce aligned span logit distillation to enhance knowledge transfer across models. In challenging cross-architecture distillation experiments, SRA consistently and significantly outperforms state-of-the-art CTKD baselines, validating our physically-grounded approach.

Quoc Phong Dao, Hoang Son Nguyen, Pham Khanh Chi +4

Hanoi University of Science and Technology · VNU University of Engineering and Technology · Monash University

Cross-Tokenizer Knowledge DistillationLanguage Model DistillationKnowledge DistillationRepresentation Alignment
AskPDF
May 1, 2026

LEAP: Layer-wise Exit-Aware Pretraining for Efficient Transformer Inference

Layer-aligned distillation and convergence-based early exit represent two predominant computational efficiency paradigms for transformer inference; yet we establish that they exhibit systematic incompatibility under standard deployment conditions for convergence-based early exit. Distillation objectives that align intermediate student layers to teacher representations suppress the representational convergence that early-exit mechanisms exploit, rendering such mechanisms ineffective on distilled models. We introduce LEAP (Layer-wise Exit-Aware Pretraining), an auxiliary training objective that reconciles this incompatibility. LEAP requires no architectural modifications; it augments standard distillation with a single constraint ensuring intermediate layers approximate final-layer representations. LEAP-MiniLM achieves 1.61×\times measured wall-clock speedup (batch=1, NVIDIA L4) at θθ=0.95, with 91.9% of samples exiting by layer 7 and 1.80×\times theoretical layer reduction, where standard distilled models achieve zero effective speedup. We validate across sentence similarity (STS-B: 0.760 ±\pm 0.006) and retrieval benchmarks (BEIR), providing operational guidance including latency measurements, decision thresholds, and deployment criteria.

Shashank Kapadia, Deep Naryan Mishra, Sujal Reddy Alugubelli +4

Walmart Inc. · Sam’s Club

Efficient Transformer InferenceDynamic Neural NetworksLanguage Model DistillationTeacher-Student LearningKnowledge DistillationEarly-Exit Neural Networks
AskPDF
May 1, 2026

AsymTalker: Identity-Consistent Long-Term Talking Head Generation via Asymmetric Distillation

Diffusion-based talking head generation has achieved remarkable visual quality, yet scaling it to long-term videos remains challenging. The widely adopted chunk-wise paradigm introduces two fundamental failures: (1) temporal-spatial misalignment between static identity references and dynamic audio streams, and (2) cascading identity drift propagated through self-generated continuity references across chunks. To address both issues, we propose AsymTalker, a novel diffusion-based talking head generation method comprising Temporal Reference Encoding (TRE) and Asymmetric Knowledge Distillation (AKD). First, TRE mitigates temporal-spatial misalignment by transforming the static identity image into a temporally coherent latent representation through encoding of a temporally replicated pseudo-video, without introducing additional parameters. Second, AKD resolves the inherent conditioning dilemma in chunk-wise training: using ground-truth references causes train-inference mismatch, while self-generated references entangle supervision with identity drift. Our asymmetric design circumvents this by anchoring the teacher model with ground-truth continuity references to provide drift-free, chunk-level supervision, thereby avoiding the teacher bottleneck. Meanwhile, the student model learns under inference-aligned conditions, conditioned only on self-generated references, and is trained via distribution matching to preserve identity over long horizons. Extensive experiments show AsymTalker achieves state-of-the-art results on HDTF and VFHQ. It guarantees high-fidelity, identity-consistent synthesis over 600-second videos and reaches a real-time inference speed of 66 FPS.

Yuxin Lu, Jiayang Sun, Guibo Zhu +1

Soochow University · Institute of Automation, Chinese Academy of Sciences

Video Diffusion ModelsLong Video GenerationTalking Head GenerationVideo GenerationKnowledge Distillation
AskPDF
May 1, 2026

Learn where to Click from Yourself: On-Policy Self-Distillation for GUI Grounding

Graphical User Interface (GUI) grounding maps natural language instructions to the visual coordinates of target elements and serves as a core capability for autonomous GUI agents. Recent reinforcement learning methods (e.g., GRPO) have achieved strong performance, but they rely on expensive multiple rollouts and suffer from sparse signals on hard samples. These limitations make on-policy self-distillation (OPSD), which provides dense token-level supervision from a single rollout, a promising alternative. However, its applicability to GUI grounding remains unexplored. In this paper, we present GUI-SD, the first OPSD framework tailored for GUI grounding. First, it constructs a visually enriched privileged context for the teacher using a target bounding box and a Gaussian soft mask, providing informative guidance without leaking exact coordinates. Second, it employs entropy-guided distillation, which adaptively weights tokens based on digit significance and teacher confidence, concentrating optimization on the most impactful and reliable positions. Extensive experiments on six representative GUI grounding benchmarks show that GUI-SD consistently outperforms GRPO-based methods and naive OPSD in both accuracy and training efficiency. Code and training data are available at https://zhangyan-ucas.github.io/GUI-SD/.

Yan Zhang, Daiqing Wu, Huawen Shen +2

Institute of Information Engineering, Chinese Academy of Sciences · School of Cyber Security, University of Chinese Academy of Sciences · VCIP & TMCC & DISSec, College of Computer Science, Nankai University

GUI GroundingOn-Policy DistillationKnowledge Distillation
AskPDF
May 1, 2026

Learning from Compressed CT: Feature Attention Style Transfer and Structured Factorized Projections for Resource-Efficient Medical Image Analysis

The deployment of artificial intelligence in medical imaging is hindered by high computational complexity and resource-intensive processing of volumetric data. Although chest computed tomography (CT) volumes offer richer diagnostic information than projection radiography, their use in AI-based diagnosis remains limited due to the computational burden of processing uncompressed volumetric images (typically stored in NIfTI or DICOM format). Addressing the growing need for low-resource deployment and efficient electronic data transfer, we investigate the utilization of JPEG-compressed chest CT volumes for thoracic abnormality detection. We propose Feature Attention Style Transfer (FAST), a novel distillation framework that transfers both activation patterns and structural relationships from high-fidelity CT representations to a spatiotemporal visual encoder operating on compressed inputs. By combining Gram-matrix-based attention style preservation with dual-attention feature alignment, FAST enables robust feature extraction from degraded volumes. Furthermore, we introduce Structured Factorized Projection (SFP), leveraging Block Tensor Train decomposition as a parameter-efficient alternative to dense projection layers, reducing projection-head parameters by almost half. Our contrastive learning pipeline, CT-Lite, integrates these components with a SigLIP-based multimodal alignment objective. Experiments on CT-RATE, NIDCH, and Rad-ChestCT demonstrate that CT-Lite achieves AUROC within 5-7% of the uncompressed-input baseline across all three datasets, despite operating on compressed inputs with significantly fewer parameters, paving the way for AI-based clinical evaluation under resource constraints.

Shadid Yousuf, S. M. Mahbubur Rahman, Mohammed Imamul Hassan Bhuiyan

Department of Electrical and Electronic Engineering, Bangladesh University of Engineering and Technology

Model CompressionMedical Image AnalysisMedical Image Anomaly DetectionComputed TomographyKnowledge Distillation
AskPDF
Apr 30, 2026

Beyond SFT-to-RL: Pre-alignment via Black-Box On-Policy Distillation for Multimodal RL

The standard post-training recipe for large multimodal models (LMMs) applies supervised fine-tuning (SFT) on curated demonstrations followed by reinforcement learning with verifiable rewards (RLVR). However, SFT introduces distributional drift that neither preserves the model's original capabilities nor faithfully matches the supervision distribution. This problem is further amplified in multimodal reasoning, where perception errors and reasoning failures follow distinct drift patterns that compound during subsequent RL. We introduce PRISM, a three-stage pipeline that mitigates this drift by inserting an explicit distribution-alignment stage between SFT and RLVR. Building on the principle of on-policy distillation (OPD), PRISM casts alignment as a black-box, response-level adversarial game between the policy and a Mixture-of-Experts (MoE) discriminator with dedicated perception and reasoning experts, providing disentangled corrective signals that steer the policy toward the supervision distribution without requiring access to teacher logits. While 1.26M public demonstrations suffice for broad SFT initialization, distribution alignment demands higher-fidelity supervision; we therefore curate 113K additional demonstrations from Gemini 3 Flash, featuring dense visual grounding and step-by-step reasoning on the hardest unsolved problems. Experiments on Qwen3-VL show that PRISM consistently improves downstream RLVR performance across multiple RL algorithms (GRPO, DAPO, GSPO) and diverse multimodal benchmarks, improving average accuracy by +4.4 and +6.0 points over the SFT-to-RLVR baseline on 4B and 8B, respectively. Our code, data, and model checkpoints are publicly available at https://github.com/XIAO4579/PRISM.

Sudong Wang, Weiquan Huang, Xiaomin Yu +9

Hong Kong University of Science and Technology (Guangzhou) · Nanyang Technological University · Tsinghua University +3

Multimodal Large Language ModelsMultimodal ReasoningOn-Policy DistillationMultimodal Foundation ModelsKnowledge DistillationRL for VLMs+1
AskPDF
Apr 29, 2026

Energy-Efficient Plant Monitoring via Knowledge Distillation

Recent advances in large-scale visual representation learning have significantly improved performance in plant species and plant disease recognition tasks. However, state-of-the-art models, often based on high-capacity vision transformers or multimodal foundation models, remain computationally expensive and difficult to deploy in resource-constrained environments such as mobile or edge devices. This limitation hinders the scalability of automated biodiversity monitoring and precision agriculture systems, where efficiency is as critical as accuracy. In this work, we investigate knowledge distillation as an effective approach to transfer the representational capacity of large pretrained models into smaller, more efficient architectures. We focus on plant species and disease recognition, and conduct an extensive empirical study on two challenging benchmarks: Pl@ntNet300K-v2 and Deep-Plant-Disease. We evaluate four representative architectures, including two ConvNeXt models and two vision transformers, under multiple training regimes: from-scratch training and pretrained initialization, each with and without distillation. In total, we train and evaluate 70 models. Our results show that knowledge distillation consistently improves performance across tasks and architectures. Distilled models are able to match the performance of significantly larger models while maintaining substantially lower computational cost. These findings demonstrate the potential of knowledge distillation techniques to enable efficient and scalable deployment of plant recognition systems in real-world environmental applications.

Ilyass Moummad, Reda Bensaid, Kawtar Zaher +5

LIRMM, Univ Montpellier, CNRS, Inria, Montpellier, France · IMT Atlantique, Brest, France. Polytechnique Montréal, Canada · Institut National de l’Audiovisuel, Paris, France +2

Plant Disease ClassificationEnergy-Efficient MLKnowledge Distillation
AskPDF
Apr 29, 2026

Lightweight Distillation of SAM 3 and DINOv3 for Edge-Deployable Individual-Level Livestock Monitoring and Longitudinal Visual Analytics

Foundation-model pipelines for individual-level livestock monitoring -- combining open-vocabulary detection, promptable video segmentation, and self-supervised visual embeddings -- have raised the accuracy ceiling of precision livestock farming (PLF), but their GPU memory budgets exceed the envelope of commodity edge accelerators. To close this gap, the 446M-parameter Perception Encoder (PE-ViT-L+) backbone of SAM 3 is distilled into a 40.66M-parameter multi-scale student through three mechanisms: a Feature Pyramid Network student encoder built on TinyViT-21M-512, a four-term direction-then-scale distillation loss, and backbone-substitution inference with sliding-window session pruning that bounds streaming GPU memory growth. The DINOv3 family includes a pre-distilled ViT-S/16 variant (21.6M parameters) released alongside a 6716M-parameter ViT-7B teacher; the ViT-S (21M) variant is adopted as the per-individual embedder. On the Edinburgh Pig dataset, the compressed pipeline reaches 92.29% MOTA and 96.15% IDF1 against the SAM 3 teacher (1.68- and 0.84-percentage-point losses), achieves a 7.77-fold reduction in system-level parameters and a 3.01-fold reduction in peak VRAM (19.52GB -> 6.49GB), and reaches 97.34% top-1 accuracy with 91.67% macro-F1 on nine-class pig behaviour classification. The pipeline fits inside an NVIDIA Jetson Orin NX 16GB envelope with 4.9GB of headroom, supporting a proposed -- but not yet empirically validated -- on-device embedding-pool re-identification mechanism whose per-individual footprint of approximately 94MB per animal per year produces a longitudinal visual record amenable to retrospective association with disease, lameness, reproductive, and growth outcome labels.

Haiyu Yang, Miel Hostens

College of Agriculture and Life Sciences, Cornell University, Ithaca, NY 14853 USA

Agricultural Image AnalysisEdge MLKnowledge DistillationMulti-Object TrackingPrecision Livestock Farming
AskPDF
Apr 29, 2026

Turning the TIDE: Cross-Architecture Distillation for Diffusion Large Language Models

Diffusion large language models (dLLMs) offer parallel decoding and bidirectional context, but state-of-the-art dLLMs require billions of parameters for competitive performance. While existing distillation methods for dLLMs reduce inference steps within a single architecture, none address cross-architecture knowledge transfer, in which the teacher and student differ in architecture, attention mechanism, and tokenizer. We present TIDE, the first framework for cross-architecture dLLM distillation, comprising three modular components: (1) TIDAL, which jointly modulates distillation strength across training progress and diffusion timestep to account for the teacher's noise-dependent reliability; (2) CompDemo, which enriches the teacher's context via complementary mask splitting to improve predictions under heavy masking; and (3) Reverse CALM, a cross-tokenizer objective that inverts chunk-level likelihood matching, yielding bounded gradients and dual-end noise filtering. Distilling 8B dense and 16B MoE teachers into a 0.6B student via two heterogeneous pipelines outperforms the baseline by an average of 1.53 points across eight benchmarks, yielding notable gains in code generation, where HumanEval scores reach 48.78 compared to 32.3 for the AR baseline.

Gongbo Zhang, Wen Wang, Ye Tian +1

Peking University · Zhejiang University

Cross-Tokenizer Knowledge DistillationDiffusion Model DistillationCross-Architecture Knowledge DistillationLanguage Model DistillationDiffusion Language ModelsKnowledge Distillation+1
AskPDF