Image de couverture de livre

Why Are DMD Students Lazy? Understanding the Copying Behavior in Few-Step Distillation

Fuente: arXiv

Enregistré dans:

Détails bibliographiques
Auteurs principaux:	Li, Shucheng, Jones, Iolo, Tong, Alexander, Bronstein, Michael M.
Format:	Preprint
Publié:	2026
Sujets:	Machine Learning
Accès en ligne:	Acceder al recurso
Tags:	Ajouter un tag Pas de tags, Soyez le premier à ajouter un tag!

Documents similaires

Carré du champ flow matching: better quality-generalisation tradeoff in generative models
par: Bamberger, Jacob, et autres
Publié: (2025)

Generalised Flow Maps for Few-Step Generative Modelling on Riemannian Manifolds
par: Davis, Oscar, et autres
Publié: (2025)

Weight Copy and Low-Rank Adaptation for Few-Shot Distillation of Vision Transformers
par: Grigore, Diana-Nicoleta, et autres
Publié: (2024)

Few-Step Diffusion Language Models via Trajectory Self-Distillation
par: Zhang, Tunyu, et autres
Publié: (2026)

Optimizing Few-Step Generation with Adaptive Matching Distillation
par: Bai, Lichen, et autres
Publié: (2026)

Few-Step Diffusion via Score identity Distillation
par: Zhou, Mingyuan, et autres
Publié: (2025)

SGMD: Score Gradient Matching Distillation for Few-Step Video Diffusion Distillation
par: Wu, Zhuguanyu, et autres
Publié: (2026)

LazyDiT: Lazy Learning for the Acceleration of Diffusion Transformers
par: Shen, Xuan, et autres
Publié: (2024)

ReDiF: Reinforced Distillation for Few Step Diffusion
par: Tighkhorshid, Amirhossein, et autres
Publié: (2025)

Factorized Neural Implicit DMD for Parametric Dynamics
par: Chen, Siyuan, et autres
Publié: (2026)

Understanding Virtual Nodes: Oversquashing and Node Heterogeneity
par: Southern, Joshua, et autres
Publié: (2024)

Privacy-Preserving Federated Learning via Dataset Distillation
par: Xu, ShiMao, et autres
Publié: (2024)

Lazy FSCA for Unsupervised Variable Selection
par: Zocco, Federico, et autres
Publié: (2021)

Trajectory as the Teacher: Few-Step Discrete Flow Matching via Energy-Navigated Distillation
par: Monsefi, Amin Karimi, et autres
Publié: (2026)

Keep Everyone Happy: Online Fair Division of Numerous Items with Few Copies
par: Verma, Arun, et autres
Publié: (2024)

Listening to the Wise Few: Select-and-Copy Attention Heads for Multiple-Choice QA
par: Tulchinskii, Eduard, et autres
Publié: (2024)

MacroGuide: Topological Guidance for Macrocycle Generation
par: Maksymiuk, Alicja, et autres
Publié: (2026)

Manifold Diffusion Geometry: Curvature, Tangent Spaces, and Dimension
par: Jones, Iolo
Publié: (2024)

Diffusion Geometry
par: Jones, Iolo
Publié: (2024)

pi-Flow: Policy-Based Few-Step Generation via Imitation Distillation
par: Chen, Hansheng, et autres
Publié: (2025)

Transformer See, Transformer Do: Copying as an Intermediate Step in Learning Analogical Reasoning
par: Hellwig, Philipp, et autres
Publié: (2026)

Demystifying Lazy Training of Neural Networks from a Macroscopic Viewpoint
par: Li, Yuqing, et autres
Publié: (2024)

AdvDMD: Adversarial Reward Meets DMD For High-Quality Few-Step Generation
par: Wang, Xu, et autres
Publié: (2026)

From Lazy to Rich: Exact Learning Dynamics in Deep Linear Networks
par: Dominé, Clémentine C. J., et autres
Publié: (2024)

DistillKac: Few-Step Image Generation via Damped Wave Equations
par: Han, Weiqiao, et autres
Publié: (2025)

Dual-Forward Path Teacher Knowledge Distillation: Bridging the Capacity Gap Between Teacher and Student
par: Li, Tong, et autres
Publié: (2025)

Partially Lazy Gradient Descent for Smoothed Online Learning
par: Mhaisen, Naram, et autres
Publié: (2026)

A Few-Step Generative Model on Cumulative Flow Maps
par: Li, Zhiqi, et autres
Publié: (2026)

John Ellipsoids via Lazy Updates
par: Woodruff, David P., et autres
Publié: (2025)

Sharpened Lazy Incremental Quasi-Newton Method
par: Lahoti, Aakash, et autres
Publié: (2023)

Efficient Learning of Quantum States Prepared With Few Non-Clifford Gates II: Single-Copy Measurements
par: Grewal, Sabee, et autres
Publié: (2023)

Scalable Equilibrium Sampling with Sequential Boltzmann Generators
par: Tan, Charlie B., et autres
Publié: (2025)

Coupling Models for One-Step Discrete Generation
par: Peng, Fred Zhangzhi, et autres
Publié: (2026)

Should Under-parameterized Student Networks Copy or Average Teacher Weights?
par: Şimşek, Berfin, et autres
Publié: (2023)

Policy Compatible Skill Incremental Learning via Lazy Learning Interface
par: Lee, Daehee, et autres
Publié: (2025)

Lazy Data Practices Harm Fairness Research
par: Simson, Jan, et autres
Publié: (2024)

Why Does Little Robustness Help? A Further Step Towards Understanding Adversarial Transferability
par: Zhang, Yechao, et autres
Publié: (2023)

FlashMol: High-Quality Molecule Generation in as Few as Four Steps
par: Wei, Xinyuan, et autres
Publié: (2026)

The Importance of Being Lazy: Scaling Limits of Continual Learning
par: Graldi, Jacopo, et autres
Publié: (2025)

BIRD: Behavior Induction via Representation-structure Distillation
par: Pogoncheff, Galen, et autres
Publié: (2025)