相关论文: Every Component is a Lookup: Token Attribution and…

Transformer architectures have achieved remarkable success across language, vision, and multimodal tasks, and there is growing demand for them to address in-context compositional learning tasks. In these tasks, models solve the target…

Standard Transformers have a fixed computational depth, fundamentally limiting their ability to generalize to tasks requiring variable-depth reasoning, such as multi-hop graph traversal or nested logic. We propose a depth-recurrent…

Decomposing model activations into interpretable components is a key open problem in mechanistic interpretability. Sparse autoencoders (SAEs) are a popular method for decomposing the internal activations of trained transformers into sparse,…

Linear probes and sparse autoencoders consistently recover meaningful structure from transformer representations -- yet why should such simple methods succeed in deep, nonlinear systems? We show this is not merely an empirical regularity…

Many machine learning tasks require finding per-part correspondences between objects. In this work we focus on low-level correspondences - a highly ambiguous matching problem. We propose to use a hierarchical semantic representation of the…

DDCL-Attention is a prototype-based readout layer for transformer encoders that replaces simple pooling methods, such as mean pooling or class tokens, with a learned compression mechanism. It uses a small set of global prototype vectors and…

Understanding internal representations of neural models is a core interest of mechanistic interpretability. Due to its large dimensionality, the representation space can encode various aspects about inputs. To what extent are different…

Locating and editing knowledge in large language models (LLMs) is crucial for enhancing their accuracy, safety, and inference rationale. We introduce ``concept editing'', an innovative variation of knowledge editing that uncovers…

Existing methods enhance the training of detection transformers by incorporating an auxiliary one-to-many assignment. In this work, we treat the model as a multi-task framework, simultaneously performing one-to-one and one-to-many…

We investigate the task of retrieving information from compositional distributed representations formed by Hyperdimensional Computing/Vector Symbolic Architectures and present novel techniques which achieve new information rate bounds.…

Mechanistic interpretability aims to understand neural networks by identifying which learned features mediate specific behaviors. Attribution graphs reveal these feature pathways, but interpreting them requires extensive manual analysis --…

Automated mechanistic interpretation research has attracted great interest due to its potential to scale explanations of neural network internals to large models. Existing automated circuit discovery work relies on activation patching or…

Patterns, which are collections of elements arranged in regular or near-regular arrangements, are an important graphic art form and widely used due to their elegant simplicity and aesthetic appeal. When a pattern is encoded as a flat image…

The concept of decomposition in computer science and engineering is considered a fundamental component of computational thinking and is prevalent in design of algorithms, software construction, hardware design, and more. We propose a simple…

To solve a new task from minimal experience, it is essential to effectively reuse knowledge from previous tasks, a problem known as meta-learning. Compositional solutions, where common elements of computation are flexibly recombined into…

Large Language Models (LLMs) have achieved impressive performance across diverse tasks but continue to struggle with learning transitive relations, a cornerstone for complex planning. To address this issue, we investigate the Multi-Token…

Through considerable effort and intuition, several recent works have reverse-engineered nontrivial behaviors of transformer models. This paper systematizes the mechanistic interpretability process they followed. First, researchers choose a…

Interpretability is a pressing issue for machine learning. Common approaches to interpretable machine learning constrain interactions between features of the input, rendering the effects of those features on a model's output comprehensible…

Understanding the internal mechanisms of large language models (LLMs) is integral to enhancing their reliability, interpretability, and inference processes. We present Constituent-Aware Pooling (CAP), a methodology designed to analyse how…

Transformers can generate predictions in two approaches: 1. auto-regressively by conditioning each sequence element on the previous ones, or 2. directly produce an output sequences in parallel. While research has mostly explored upon this…