English
Related papers

Related papers: When LRP Diverges from Leave-One-Out in Transforme…

200 papers

In recent decades, multilevel regression and poststratification (MRP) has surged in popularity for population inference. However, the validity of the estimates can depend on details of the model, and there is currently little research on…

Methodology · Statistics 2022-09-07 Swen Kuh , Lauren Kennedy , Qixuan Chen , Andrew Gelman

While scaling Transformer-based large language models (LLMs) has demonstrated promising performance across various tasks, it also introduces redundant architectures, posing efficiency challenges for real-world deployment. Despite some…

Machine Learning · Computer Science 2024-10-18 Shwai He , Guoheng Sun , Zheyu Shen , Ang Li

Large Language Models (LLMs) have shown immense potential in enhancing various aspects of our daily lives, from conversational AI to search and AI assistants. However, their growing capabilities come at the cost of extremely large model…

Machine Learning · Computer Science 2025-02-27 Yingyu Liang , Jiangxuan Long , Zhenmei Shi , Zhao Song , Yufa Zhou

Transformers are mostly relying on softmax attention, which introduces quadratic complexity with respect to sequence length and remains a major bottleneck for efficient inference. Prior work on linear or hybrid attention typically replaces…

Large language models (LLMs) have gained much attention in the recommendation community; some studies have observed that LLMs, fine-tuned by the cross-entropy loss with a full softmax, could achieve state-of-the-art performance already.…

Information Retrieval · Computer Science 2024-02-23 Cong Xu , Zhangchi Zhu , Jun Wang , Jianyong Wang , Wei Zhang

Reinforcement learning (RL)-based fine-tuning has emerged as a powerful approach for aligning diffusion models with black-box objectives. Proximal policy optimization (PPO) is a popular choice of method for policy optimization. While…

Compressing transformer weights makes large language models cheaper to deploy. But each layer's compression introduces an error. These errors accumulate as the signal passes through later layers, and how they accumulate is not well…

Machine Learning · Computer Science 2026-05-08 Abhinaba Basu , Kumkum Basu , Koushik Deb

Sequential decision-making algorithms such as reinforcement learning (RL) in real-world scenarios inevitably face environments with partial observability. This paper scrutinizes the effectiveness of a popular architecture, namely…

Machine Learning · Computer Science 2024-05-31 Chenhao Lu , Ruizhe Shi , Yuyao Liu , Kaizhe Hu , Simon S. Du , Huazhe Xu

We study how information propagates in decoder-only Transformers, which are the architectural backbone of most existing frontier large language models (LLMs). We rely on a theoretical signal propagation analysis -- specifically, we analyse…

Artificial neural networks (ANNs) are known to be powerful methods for many hard problems (e.g. image classification, speech recognition or time series prediction). However, these models tend to produce black-box results and are often…

Machine Learning · Computer Science 2022-11-17 Marco Landt-Hayen , Peer Kröger , Martin Claus , Willi Rath

In contrast to RNNs, which compress their history into a single hidden state, Transformers can attend to all past tokens directly. However, standard Transformers rely solely on the hidden state from the previous layer to represent the…

Machine Learning · Computer Science 2025-05-29 Gleb Gerasimov , Yaroslav Aksenov , Nikita Balagansky , Viacheslav Sinii , Daniil Gavrilov

Learning rate configuration is a fundamental aspect of modern deep learning. The prevailing practice of applying a uniform learning rate across all layers overlooks the structural heterogeneity of Transformers, potentially limiting their…

Machine Learning · Computer Science 2026-05-28 Di He , Songjun Tu , Keyu Wang , Lu Yin , Shiwei Liu

Activation patching is a standard method in mechanistic interpretability for localizing the components of a model responsible for specific behaviors, but it is computationally expensive to apply at scale. Attribution patching offers a…

Machine Learning · Computer Science 2025-10-31 Farnoush Rezaei Jafari , Oliver Eberle , Ashkan Khakzar , Neel Nanda

Large Language Models (LLMs) often exhibit a gap between their internal knowledge and their explicit linguistic outputs. In this report, we empirically investigate whether Looped Transformers (LTs)--architectures that increase computational…

Computation and Language · Computer Science 2026-01-16 Guanxu Chen , Dongrui Liu , Jing Shao

Despite their empirical success, the internal mechanism by which transformer models align tokens during language processing remains poorly understood. This paper provides a mechanistic and theoretical explanation of token alignment in LLMs.…

Machine Learning · Computer Science 2025-12-19 Hadi Daneshmand

Preference alignment is pivotal for empowering large language models (LLMs) to generate helpful and harmless responses. However, the performance of preference alignment is highly sensitive to the prevalent noise in the preference data.…

Machine Learning · Computer Science 2024-05-29 Xize Liang , Chao Chen , Shuang Qiu , Jie Wang , Yue Wu , Zhihang Fu , Zhihao Shi , Feng Wu , Jieping Ye

Representation Misdirection for Unlearning (RMU), which steers model representation in the intermediate layer to a target random representation, is an effective method for large language model (LLM) unlearning. Despite its high performance,…

Computation and Language · Computer Science 2025-02-07 Dang Huu-Tien , Trung-Tin Pham , Hoang Thanh-Tung , Naoya Inoue

We show that MLP layers in transformer language models perform binary routing of continuous signals: the decision of whether a token needs nonlinear processing is well-captured by binary neuron activations, even though the signals being…

Machine Learning · Computer Science 2026-03-12 Peter Balogh

In this paper we generalize and extend an idea of low-rank adaptation (LoRA) of large language models (LLMs) based on Transformer architecture. Widely used LoRA-like methods of fine-tuning LLMs are based on matrix factorization of gradient…

Computation and Language · Computer Science 2024-02-06 Daniel Bershatsky , Daria Cherniuk , Talgat Daulbaev , Aleksandr Mikhalev , Ivan Oseledets

Attribution methods aim to explain a neural network's prediction by highlighting the most relevant image areas. A popular approach is to backpropagate (BP) a custom relevance score using modified rules, rather than the gradient. We analyze…

Machine Learning · Computer Science 2024-02-20 Leon Sixt , Maximilian Granz , Tim Landgraf