Image de couverture de livre

PV-Tuning: Beyond Straight-Through Estimation for Extreme LLM Compression

Fuente: arXiv

Enregistré dans:

Détails bibliographiques
Auteurs principaux:	Malinovskii, Vladimir, Mazur, Denis, Ilin, Ivan, Kuznedelev, Denis, Burlachenko, Konstantin, Yi, Kai, Alistarh, Dan, Richtarik, Peter
Format:	Preprint
Publié:	2024
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

Cache Me If You Must: Adaptive Key-Value Quantization for Large Language Models
par: Shutova, Alina, et autres
Publié: (2025)

Pushing the Limits of Large Language Model Quantization via the Linearity Theorem
par: Malinovskii, Vladimir, et autres
Publié: (2024)

EvoPress: Accurate Dynamic Model Compression via Evolutionary Search
par: Sieberling, Oliver, et autres
Publié: (2024)

Extreme Compression of Large Language Models via Additive Quantization
par: Egiazarian, Vage, et autres
Publié: (2024)

Unlocking FedNL: Self-Contained Compute-Optimized Implementation
par: Burlachenko, Konstantin, et autres
Publié: (2024)

BurTorch: Revisiting Training from First Principles by Coupling Autodiff, Math Optimization, and Systems
par: Burlachenko, Konstantin, et autres
Publié: (2025)

Error Feedback Reloaded: From Quadratic to Arithmetic Mean of Smoothness Constants
par: Richtárik, Peter, et autres
Publié: (2024)

Thanos: A Block-wise Pruning Algorithm for Efficient Large Language Model Compression
par: Ilin, Ivan, et autres
Publié: (2025)

The Iterative Optimal Brain Surgeon: Faster Sparse Recovery by Leveraging Second-Order Information
par: Wu, Diyuan, et autres
Publié: (2024)

Optimization Methods and Software for Federated Learning
par: Burlachenko, Konstantin
Publié: (2025)

Hogwild! Inference: Parallel LLM Generation via Concurrent Attention
par: Rodionov, Gleb, et autres
Publié: (2025)

Characterizing Graph Datasets for Node Classification: Homophily-Heterophily Dichotomy and Beyond
par: Platonov, Oleg, et autres
Publié: (2022)

Model Compression with Exact Budget Constraints via Riemannian Manifolds
par: Helcig, Michael, et autres
Publié: (2026)

Shadowheart SGD: Distributed Asynchronous SGD with Optimal Time Complexity Under Arbitrary Computation and Communication Heterogeneity
par: Tyurin, Alexander, et autres
Publié: (2024)

Accurate Compression of Text-to-Image Diffusion Models via Vector Quantization
par: Egiazarian, Vage, et autres
Publié: (2024)

Ringmaster LMO: Asynchronous Linear Minimization Oracle Momentum Method
par: Sadiev, Abdurakhmon, et autres
Publié: (2026)

Symmetric Pruning of Large Language Models
par: Yi, Kai, et autres
Publié: (2025)

KV Cache Offloading for Context-Intensive Tasks
par: Bocharnikov, Andrey, et autres
Publié: (2026)

Cohort Squeeze: Beyond a Single Communication Round per Cohort in Cross-Device Federated Learning
par: Yi, Kai, et autres
Publié: (2024)

MatryoshkaLoRA: Learning Accurate Hierarchical Low-Rank Representations for LLM Fine-Tuning
par: Modoranu, Ionut-Vlad, et autres
Publié: (2026)

Custom Gradient Estimators are Straight-Through Estimators in Disguise
par: Schoenbauer, Matt, et autres
Publié: (2024)

Quartet II: Accurate LLM Pre-Training in NVFP4 by Improved Unbiased Gradient Estimation
par: Panferov, Andrei, et autres
Publié: (2026)

Bridging the Gap Between Promise and Performance for Microscaling FP4 Quantization
par: Egiazarian, Vage, et autres
Publié: (2025)

A critical look at the evaluation of GNNs under heterophily: Are we really making progress?
par: Platonov, Oleg, et autres
Publié: (2023)

CompressedScaffnew: The First Theoretical Double Acceleration of Communication from Local Training and Compression in Distributed Optimization
par: Condat, Laurent, et autres
Publié: (2022)

Improving the Straight-Through Estimator with Zeroth-Order Information
par: Yang, Ningfeng, et autres
Publié: (2025)

TAMUNA: Doubly Accelerated Distributed Optimization with Local Training, Compression, and Partial Participation
par: Condat, Laurent, et autres
Publié: (2023)

MicroAdam: Accurate Adaptive Optimization with Low Space Overhead and Provable Convergence
par: Modoranu, Ionut-Vlad, et autres
Publié: (2024)

Behemoth: Benchmarking Unlearning in LLMs Using Fully Synthetic Data
par: Iofinova, Eugenia, et autres
Publié: (2026)

Improving Discrete Optimisation Via Decoupled Straight-Through Estimator
par: Shah, Rushi, et autres
Publié: (2024)

Apertus LLM Family Expansion via Distillation and Quantization
par: Panferov, Andrei, et autres
Publié: (2026)

Compression Scaling Laws:Unifying Sparsity and Quantization
par: Frantar, Elias, et autres
Publié: (2025)

Beyond Discreteness: Sample Complexity Analysis of Straight-Through Estimator for 1-bit Quantization
par: Jeong, Halyun, et autres
Publié: (2025)

RoSA: Accurate Parameter-Efficient Fine-Tuning via Robust Adaptation
par: Nikdan, Mahdi, et autres
Publié: (2024)

BiCoLoR: Communication-Efficient Optimization with Bidirectional Compression and Local Training
par: Condat, Laurent, et autres
Publié: (2026)

Beyond Outliers: A Study of Optimizers Under Quantization
par: Vlassis, Georgios, et autres
Publié: (2025)

Label Privacy in Split Learning for Large Models with Parameter-Efficient Training
par: Zmushko, Philip, et autres
Publié: (2024)

LLMQ: Efficient Lower-Precision Pretraining for Consumer GPUs
par: Schultheis, Erik, et autres
Publié: (2025)

Sparse-ProxSkip: Accelerated Sparse-to-Sparse Training in Federated Learning
par: Meinhardt, Georg, et autres
Publié: (2024)

FedP3: Federated Personalized and Privacy-friendly Network Pruning under Model Heterogeneity
par: Yi, Kai, et autres
Publié: (2024)