18 papers · 1 filter
Not All Drift Is a Mechanism: Tracing Prompt Adaptation in LLMs
Mikhail Aleksandrov, Nail Bashirov, Maksim Ivanov +6
Prompt-based adaptation offers several ways to specialize a frozen language model, ranging from natural-language prompt optimization to learned continuous prompts. These methods ar…
Multi-Stream Residual Computation in Diffusion Language Models
Ekaterina Alimaskina, Suren Kazarian, Gleb Molodtsov +1
Diffusion language models reconstruct masked tokens in parallel, offering an alternative to autoregressive generation. Yet their Transformer backbones typically carry visible evide…
Monarch-Muon: Structure-Aware Muon for Fast and Memory-Efficient Pretraining
Alexander Miasnikov, Ruslan Kabirov, Gleb Molodtsov +7
Training large language models is constrained by both wall-clock time and memory footprint, and recent matrix-aware optimizers sharpen this tradeoff: Muon can improve optimization…
Mask-Guided KV Cache Eviction in Block Diffusion Language Models
Gleb Molodtsov, Ekaterina Alimaskina, Evgeny Uskov +2
Block diffusion language models keep a large key-value (KV) cache throughout generation and attend to it at every denoising step, limiting both memory capacity and generation speed…
LionMuon: Alternating Spectral and Sign Descent for Efficient Training
Arman Bolatov, Artem Riabinin, Nikita Kornilov +4
Pretraining a language model takes enormous compute, and the right optimizer can save a good part of it. Muon's spectral step gives a stronger direction than a sign step, but it is…
LoRA-TSD: Tangent-Space Spectral Descent for LoRA via Muon-Style Updates
Dmitrii Andriianov, Andrey Veprikov, Aleksandr Beznosikov
Low-rank adaptation (LoRA) is the standard way to fine-tune large models, yet when its two factors are trained independently, the update ignores the geometry of the low-rank weight…
Controlling Refusal Behavior of LLMs via Stiefel-Constrained Rotation Steering
Kirill Bunin, Dmitry Bylinkin, Vladimir Aletov +3
Activation steering has emerged as a lightweight approach for controlling model refusal at inference time. A growing line of research explores trainable rotations of activations to…
Leveraging Association Context Retrieval in Knowledge Edit- ing to Build White-Box Attacks on LLMs
Roman Maksimov, Vladimir Aletov, Vladimir Solodkin +3
As large language models (LLMs) are granted increasing autonomy, it is essential to investigate methods that can induce unsafe behavior. We propose a novel white-box attack inspire…
Zero-order Parameter-free Optimization for LMO-based Methods: Novel Approach for Efficient Fine-tuning
Dmitriy Bystrov, Daniil Medyakov, Dmitry Bylinkin +1
Fine-tuning large language models (LLMs) has become a central application of modern optimization, enabling pretrained models to adapt to diverse downstream tasks and domain-specifi…
Analyzing Stream Collapse in Hyper-Connections: From Diagnosis to Mitigation
Ekaterina Alimaskina, Gleb Molodtsov, Aleksandr Beznosikov
Hyper-Connections (HC) replace the single Transformer residual stream with multiple streams, introducing a permutation symmetry over stream indices. We study how this symmetry is r…
Rethinking the Role of Tensor Decompositions in Post-Training LLM Compression
Artur Zagitov, Alexander Miasnikov, Maxim Krutikov +5
Post-training compression is essential for deploying large language models (LLMs) under tight resource constraints. Tensor decompositions have emerged as a promising direction, off…
Softsign: Smooth Sign in Your Optimizer For Better Parameter Heterogeneity Handling
Dmitrii Feoktistov, Timofey Belinsky, Andrey Veprikov +2
Sign-based and LMO-inspired optimizers have recently attracted substantial attention in deep learning due to their strong performance and low memory footprint. However, their fixed…
Why SGD is not Brownian Motion: A New Perspective on Stochastic Dynamics
Igor Ignashin, Anna Radovskaya, Andrew Semenov +7
Stochastic Gradient Descent (SGD) is commonly modeled as a Langevin process, assuming that minibatch noise acts as Brownian motion. However, this approximation relies on a continuo…
Scalable Knowledge Editing for Mixture-of-Experts LLMs via Tensor-Structured Updates
Roman Maksimov, Vladimir Aletov, Dmitry Bylinkin +3
Knowledge editing (KE) provides a lightweight alternative to repeated fine-tuning of LLMs. However, most existing KE methods target dense feed-forward layers, while modern LLMs inc…
Hierarchical Mixture-of-Experts with Two-Stage Optimization
Gleb Molodtsov, Alexander Miasnikov, Aleksandr Beznosikov
Sparse Mixture-of-Experts (MoE) models scale capacity by routing each token to a small subset of experts. However, their routers exhibit a fundamental trade-off: strong load balanc…
HARP: Hadamard-Preconditioned Adaptive Rotation Processor for Extreme LLM Quantization
Artur Zagitov, Gleb Molodtsov, Aleksandr Beznosikov
Post-training quantization (PTQ) is essential for deploying LLMs under memory and bandwidth constraints. However, extreme low-bit quantization remains highly sensitive to activatio…
LionMuon: Alternating Spectral and Sign Descent for Efficient Training
Arman Bolatov, Artem Riabinin, Nikita Kornilov +4
In large-scale optimization, the cheapness and effectiveness of update steps are the most crucial factors for a successful optimizer. Sign-based optimizers like Lion or Signum prod…
Zero-Order Optimization for LLM Fine-Tuning via Learnable Direction Sampling
Valery Parfenov, Grigoriy Evseev, Andrey Veprikov +3
Fine-tuning large pretrained language models (LLMs) is a cornerstone of modern NLP, yet its growing memory demands (driven by backpropagation and large optimizer States) limit depl…