English
Related papers

Related papers: Toy Models of Superposition

200 papers

Model interpretability methods are often used to explain NLP model decisions on tasks such as text classification, where the output space is relatively small. However, when applied to language generation, where the output space often…

Computation and Language · Computer Science 2022-05-24 Kayo Yin , Graham Neubig

Sentence embeddings are central to modern NLP and AI systems, yet little is known about their internal structure. While we can compare these embeddings using measures such as cosine similarity, the contributing features are not…

Computation and Language · Computer Science 2025-06-11 Matthieu Tehenan , Vikram Natarajan , Jonathan Michala , Milton Lin , Juri Opitz

In order to better understand a complex theory like quantum mechanics, it is sometimes useful to take a step back and create alternative theories, with more intuitive foundations, and examine which features of quantum mechanics can be…

Quantum Physics · Physics 2021-05-10 Ladina Hausmann , Nuriya Nurgalieva , Lídia del Rio

If the same neuron activates for both "lender" and "riverside," standard metrics attribute the overlap to superposition--the neuron must be compressing two unrelated concepts. This work explores how much of the overlap is due a lexical…

Computation and Language · Computer Science 2026-04-02 Iyad Ait Hou , Rebecca Hwa

Neural networks are capable of superposition -- representing more features than there are dimensions. Recent work considers the analogous concept for computation instead of storage, proposing theoretical constructions. But there has been…

Machine Learning · Computer Science 2025-07-15 Adam Newgas

Replay in the brain has been viewed as rehearsal, or, more recently, as sampling from a transition model. Here, we propose a new hypothesis: that replay is able to implement a form of compositional computation where entities are assembled…

Neurons and Cognition · Quantitative Biology 2022-12-21 Zeb Kurth-Nelson , Timothy Behrens , Greg Wayne , Kevin Miller , Lennart Luettgau , Ray Dolan , Yunzhe Liu , Philipp Schwartenbeck

Words can have multiple senses. Compositional distributional models of meaning have been argued to deal well with finer shades of meaning variation known as polysemy, but are not so well equipped to handle word senses that are…

Computation and Language · Computer Science 2020-10-13 Francois Meyer , Martha Lewis

We propose a new interpretability method for neural networks, which is based on a novel mathematico-philosophical theory of reasons. Our method computes a vector for each neuron, called its reasons vector. We then can compute how strongly…

Machine Learning · Computer Science 2025-05-21 Levin Hornischer , Hannes Leitgeb

Lexical ambiguity presents a profound and enduring challenge to the language sciences. Researchers for decades have grappled with the problem of how language users learn, represent and process words with more than one meaning. Our work…

Computation and Language · Computer Science 2023-04-27 Benedetta Cevoli , Chris Watkins , Yang Gao , Kathleen Rastle

One of the central aspects of contextualised language models is that they should be able to distinguish the meaning of lexically ambiguous words by their contexts. In this paper we investigate the extent to which the contextualised…

Computation and Language · Computer Science 2021-09-30 Janosch Haber , Massimo Poesio

Polysemanticity is pervasive in language models and remains a major challenge for interpretation and model behavioral control. Leveraging sparse autoencoders (SAEs), we map the polysemantic topology of two small models (Pythia-70M and…

Artificial Intelligence · Computer Science 2026-03-19 Bofan Gong , Shiyang Lai , James Evans , Dawn Song

Along with the great success of deep neural networks, there is also growing concern about their black-box nature. The interpretability issue affects people's trust on deep learning systems. It is also related to many ethical problems, e.g.,…

Machine Learning · Computer Science 2022-02-01 Yu Zhang , Peter Tiňo , Aleš Leonardis , Ke Tang

We describe a procedure for explaining neurons in deep representations by identifying compositional logical concepts that closely approximate neuron behavior. Compared to prior work that uses atomic labels as explanations, analyzing neurons…

Machine Learning · Computer Science 2021-02-04 Jesse Mu , Jacob Andreas

Machine translation is a popular test bed for research in neural sequence-to-sequence models but despite much recent research, there is still a lack of understanding of these models. Practitioners report performance degradation with large…

Computation and Language · Computer Science 2018-08-14 Myle Ott , Michael Auli , David Grangier , Marc'Aurelio Ranzato

We describe an "interpretability illusion" that arises when analyzing the BERT model. Activations of individual neurons in the network may spuriously appear to encode a single, simple concept, when in fact they are encoding something far…

Computation and Language · Computer Science 2021-04-16 Tolga Bolukbasi , Adam Pearce , Ann Yuan , Andy Coenen , Emily Reif , Fernanda Viégas , Martin Wattenberg

The unification of low-level perception and high-level reasoning is a long-standing problem in artificial intelligence, which has the potential to not only bring the areas of logic and learning closer together but also demonstrate how…

Artificial Intelligence · Computer Science 2019-11-27 Anton Fuxjaeger , Vaishak Belle

How universal is human conceptual structure? The way concepts are organized in the human brain may reflect distinct features of cultural, historical, and environmental background in addition to properties universal to human cognition.…

In traditional Distributional Semantic Models (DSMs) the multiple senses of a polysemous word are conflated into a single vector space representation. In this work, we propose a DSM that learns multiple distributional representations of a…

Computation and Language · Computer Science 2019-04-12 Eleftheria Briakou , Nikos Athanasiou , Alexandros Potamianos

A longstanding question in cognitive science concerns the learning mechanisms underlying compositionality in human cognition. Humans can infer the structured relationships (e.g., grammatical rules) implicit in their sensory observations…

Machine Learning · Computer Science 2021-05-20 Jacob Russin , Roland Fernandez , Hamid Palangi , Eric Rosen , Nebojsa Jojic , Paul Smolensky , Jianfeng Gao

The polysemantic nature of synthetic neurons in artificial intelligence language models is currently understood as the result of a necessary superposition of distributed features within the latent space. We propose an alternative approach,…