Cover Image

Zero-Order Optimization for LLM Fine-Tuning via Learnable Direction Sampling

Fuente: arXiv

Saved in:

Bibliographic Details
Main Authors:	Parfenov, Valery, Evseev, Grigoriy, Veprikov, Andrey, Bushkov, Nikolay, Moiseev, Stanislav, Beznosikov, Aleksandr
Format:	Preprint
Published:	2026
Subjects:	Machine Learning Optimization and Control
Online Access:	Acceder al recurso
Tags:	Add Tag No Tags, Be the first to tag this record!

Similar Items

Leveraging Coordinate Momentum in SignSGD and Muon: Memory-Optimized Zero-Order
by: Petrov, Egor, et al.
Published: (2025)

Shuffling Heuristic in Variational Inequalities: Establishing New Convergence Guarantees
by: Medyakov, Daniil, et al.
Published: (2025)

Sign-SGD via Parameter-Free Optimization
by: Medyakov, Daniil, et al.
Published: (2025)

Where Does Warm-Up Come From? Adaptive Scheduling for Norm-Constrained Optimizers
by: Riabinin, Artem, et al.
Published: (2026)

WeightLoRA: Keep Only Necessary Adapters
by: Veprikov, Andrey, et al.
Published: (2025)

Unified Theory of Adaptive Variance Reduction
by: Shestakov, Aleksandr, et al.
Published: (2025)

New Aspects of Black Box Conditional Gradient: Variance Reduction and One Point Feedback
by: Veprikov, Andrey, et al.
Published: (2024)

Methods for Optimization Problems with Markovian Stochasticity and Non-Euclidean Geometry
by: Solodkin, Vladimir, et al.
Published: (2024)

Preconditioned Norms: A Unified Framework for Steepest Descent, Quasi-Newton and Adaptive Methods
by: Veprikov, Andrey, et al.
Published: (2025)

Ito Diffusion Approximation of Universal Ito Chains for Sampling, Optimization and Boosting
by: Ustimenko, Aleksei, et al.
Published: (2023)

Markovian Compression: Looking to the Past Helps Accelerate the Future
by: Veprikov, Andrey, et al.
Published: (2026)

DyKAF: Dynamical Kronecker Approximation of the Fisher Information Matrix for Gradient Preconditioning
by: Yudin, Nikolay, et al.
Published: (2025)

Sarah Frank-Wolfe: Methods for Constrained Optimization with Best Rates and Practical Features
by: Beznosikov, Aleksandr, et al.
Published: (2023)

Random-reshuffled SARAH does not need a full gradient computations
by: Beznosikov, Aleksandr, et al.
Published: (2021)

Optimal Data Splitting in Distributed Optimization for Machine Learning
by: Medyakov, Daniil, et al.
Published: (2024)

Accelerated Methods with Compressed Communications for Distributed Optimization Problems under Data Similarity
by: Bylinkin, Dmitry, et al.
Published: (2024)

Hierarchical Mixture-of-Experts with Two-Stage Optimization
by: Molodtsov, Gleb, et al.
Published: (2026)

Convergence of Clipped-SGD for Convex $(L_0,L_1)$-Smooth Optimization with Heavy-Tailed Noise
by: Chezhegov, Savelii, et al.
Published: (2025)

Accelerated Stochastic ExtraGradient: Mixing Hessian and Gradient Similarity to Reduce Communication in Distributed and Federated Learning
by: Bylinkin, Dmitry, et al.
Published: (2024)

First Order Methods with Markovian Noise: from Acceleration to Variational Inequalities
by: Beznosikov, Aleksandr, et al.
Published: (2023)

Enhancing Stability of Physics-Informed Neural Network Training Through Saddle-Point Reformulation
by: Bylinkin, Dmitry, et al.
Published: (2025)

Gradient-Free Approaches is a Key to an Efficient Interaction with Markovian Stochasticity
by: Prokhorov, Boris, et al.
Published: (2026)

Extragradient Sliding for Composite Non-Monotone Variational Inequalities
by: Emelyanov, Roman, et al.
Published: (2024)

Local SGD for Near-Quadratic Problems: Improving Convergence under Unconstrained Noise Conditions
by: Sadchikov, Andrey, et al.
Published: (2024)

Variance Reduction Methods Do Not Need to Compute Full Gradients: Improved Efficiency through Shuffling
by: Medyakov, Daniil, et al.
Published: (2025)

The Black-Box Optimization Problem: Zero-Order Accelerated Stochastic Method via Kernel Approximation
by: Lobanov, Aleksandr, et al.
Published: (2023)

Power of Generalized Smoothness in Stochastic Convex Optimization: First- and Zero-Order Algorithms
by: Lobanov, Aleksandr, et al.
Published: (2025)

Distributed Saddle-Point Problems: Lower Bounds, Near-Optimal and Robust Algorithms
by: Beznosikov, Aleksandr, et al.
Published: (2020)

Effective Method with Compression for Distributed and Federated Cocoercive Variational Inequalities
by: Medyakov, Daniil, et al.
Published: (2024)

Accelerated Methods with Complexity Separation Under Data Similarity for Federated Learning Problems
by: Bylinkin, Dmitry, et al.
Published: (2026)

Symmetric Linear Dynamical Systems are Learnable from Few Observations
by: Vu, Minh, et al.
Published: (2025)

Online Optimization Perspective on First-Order and Zero-Order Decentralized Nonsmooth Nonconvex Stochastic Optimization
by: Sahinoglu, Emre, et al.
Published: (2024)

Accelerated Zero-Order SGD Method for Solving the Black Box Optimization Problem under "Overparametrization" Condition
by: Lobanov, Aleksandr, et al.
Published: (2023)

Efficient Tail-Aware Generative Optimization via Flow Model Fine-Tuning
by: Wang, Zifan, et al.
Published: (2026)

Stochastic Gradient Methods with Preconditioned Updates
by: Sadiev, Abdurakhmon, et al.
Published: (2022)

Exploring New Frontiers in Vertical Federated Learning: the Role of Saddle Point Reformulation
by: Beznosikov, Aleksandr, et al.
Published: (2026)

Bant: Byzantine Antidote via Trial Function and Trust Scores
by: Molodtsov, Gleb, et al.
Published: (2025)

Activations and Gradients Compression for Model-Parallel Training
by: Rudakov, Mikhail, et al.
Published: (2024)

On Biased Compression for Distributed Learning
by: Beznosikov, Aleksandr, et al.
Published: (2020)

Decentralized Personalized Federated Learning for Min-Max Problems
by: Borodich, Ekaterina, et al.
Published: (2021)