English
Related papers

Related papers: Engineering Verifiable Modularity in Transformers …

200 papers

Interpretability research often aims to predict how a model will respond to targeted interventions on specific mechanisms. However, it rarely predicts how a model will respond to unseen input data. This paper explores the promises and…

Machine Learning · Computer Science 2025-07-10 Victoria R. Li , Jenny Kaufmann , Martin Wattenberg , David Alvarez-Melis , Naomi Saphra

Although Large Vision-Language Models (LVLMs) have made substantial progress, hallucination, where generated text is not grounded in the visual input, remains a challenge. As LVLMs become stronger, previously reported hallucination…

Computer Vision and Pattern Recognition · Computer Science 2026-03-03 April Fu

The multi-head self-attention mechanism of the transformer model has been thoroughly investigated recently. In one vein of study, researchers are interested in understanding why and how transformers work. In another vein, researchers…

Computation and Language · Computer Science 2022-10-28 Raymond Li , Wen Xiao , Linzi Xing , Lanjun Wang , Gabriel Murray , Giuseppe Carenini

The attention mechanism in its standard implementation contains extraneous rotational degrees of freedom that are carried through computation but do not affect model activations or outputs. We introduce a simple symmetry-breaking protocol…

Machine Learning · Computer Science 2026-02-13 Eva Silverstein , Daniel Kunin , Vasudev Shyam

Vision Transformers (ViTs) have achieved state-of-the-art performance in image classification, yet their attention mechanisms often remain opaque and exhibit dense, non-structured behaviors. In this work, we adapt our previously proposed…

Computer Vision and Pattern Recognition · Computer Science 2026-02-12 Vasileios Arampatzakis , George Pavlidis , Nikolaos Mitianoudis , Nikos Papamarkos

We introduce directional routing, a lightweight mechanism that gives each transformer attention head learned suppression directions controlled by a shared router, at 3.9% parameter cost. We train a 433M-parameter model alongside an…

Machine Learning · Computer Science 2026-03-17 Kevin Taylor

Transformer-based language models exhibit complex and distributed behavior, yet their internal computations remain poorly understood. Existing mechanistic interpretability methods typically treat attention heads and multilayer perceptron…

Machine Learning · Computer Science 2025-11-26 Areeb Ahmad , Abhinav Joshi , Ashutosh Modi

Transformer models typically calculate attention matrices using dot products, which have limitations when capturing nonlinear relationships between embedding vectors. We propose Neural Attention, a technique that replaces dot products with…

Machine Learning · Computer Science 2025-11-10 Andrew DiGiugno , Ausif Mahmood

Learning algorithms become more powerful, often at the cost of increased complexity. In response, the demand for algorithms to be transparent is growing. In NLP tasks, attention distributions learned by attention-based deep learning models…

Computation and Language · Computer Science 2019-07-09 Joris Baan , Maartje ter Hoeve , Marlies van der Wees , Anne Schuth , Maarten de Rijke

The self-attention mechanism, a cornerstone of Transformer-based state-of-the-art deep learning architectures, is largely heuristic-driven and fundamentally challenging to interpret. Establishing a robust theoretical foundation to explain…

Computer Vision and Pattern Recognition · Computer Science 2026-02-10 Laziz U. Abdullaev , Maksim Tkachenko , Tan M. Nguyen

Transformers' quadratic computational complexity limits their scalability despite remarkable performance. While linear attention reduces this to linear complexity, pre-training such models from scratch remains, in most cases, prohibitively…

Machine Learning · Computer Science 2025-10-13 Martin Benfeghoul , Teresa Delgado , Adnan Oomerjee , Haitham Bou Ammar , Jun Wang , Zafeirios Fountas

Mechanistic interpretability has identified small sets of attention heads that implement specific behaviours in transformer language models, but recovering these circuits typically requires a bespoke analytical pipeline for each new task.…

Machine Learning · Computer Science 2026-05-27 Barsat Khadka

Pre-trained Transformers often exhibit over-confidence in source patterns and difficulty in forming new target-domain patterns during fine-tuning. We formalize the mechanism of output saturation leading to gradient suppression through…

Machine Learning · Computer Science 2025-11-04 Wang Zixian

Language-conditioned policies allow robots to interpret and execute human instructions. Learning such policies requires a substantial investment with regards to time and compute resources. Still, the resulting controllers are highly…

Robotics · Computer Science 2022-12-12 Yifan Zhou , Shubham Sonawani , Mariano Phielipp , Simon Stepputtis , Heni Ben Amor

Transformers enable powerful content-based global routing via self-attention, but they lack an explicit local geometric prior along the sequence axis. As a result, the placement of locality-inducing modules in hybrid architectures has…

Machine Learning · Computer Science 2026-02-17 Yukun Zhang , Xueqing Zhou

Multimodal large reasoning models (MLRMs) often suffer from hallucinations that stem not only from insufficient visual grounding but also from imbalanced allocation between perception and reasoning processes. Building upon recent…

Artificial Intelligence · Computer Science 2026-03-10 Haolang Lu , Bolun Chu , WeiYe Fu , Guoshun Nan , Junning Liu , Minghui Pan , Qiankun Li , Yi Yu , Hua Wang , Kun Wang

We propose a geometric taxonomy of large language model hallucinations based on observable signatures in token embedding cluster structure. By analyzing the static embedding spaces of 11 transformer models spanning encoder (BERT, RoBERTa,…

Computation and Language · Computer Science 2026-02-17 Matic Korun

Transformers trained on modular arithmetic exhibit sharp transitions between memorization, generalization, and collapse. We show that weight decay acts as a scalar empirical control parameter for these regimes, and introduce two cheap…

Machine Learning · Computer Science 2026-05-21 Lucky Verma

A widely cited result by Dong et al. (2021) showed that Transformers built from self-attention alone, without skip connections or feed-forward layers, suffer from rapid rank collapse: all token representations converge to a single…

Machine Learning · Computer Science 2026-04-28 Giansalvo Cirrincione

Architectural obfuscation - e.g., permuting hidden-state tensors, linearly transforming embedding tables, or remapping tokens - has recently gained traction as a lightweight substitute for heavyweight cryptography in privacy-preserving…

Cryptography and Security · Computer Science 2025-06-24 Marcos Florencio , Thomas Barton