papers

Publications (50)

cs.CL2022

Causal Analysis of Syntactic Agreement Neurons in Multilingual Language Models

Aaron Mueller, Yu Xia, Tal Linzen

Structural probing work has found evidence for latent syntactic information in pre-trained language models. However, much of this analysis has focused on monolingual models, and an…

cs.LG2026

Causality is Key for Interpretability Claims to Generalise

Shruti Joshi, Aaron Mueller, David Klindt +3

Interpretability research on large language models (LLMs) has yielded important insights into model behaviour, yet recurring pitfalls persist: findings that do not generalise, and…

cs.LG2025

SAEs Are Good for Steering -- If You Select the Right Features

Dana Arad, Aaron Mueller, Yonatan Belinkov

Sparse Autoencoders (SAEs) have been proposed as an unsupervised approach to learn a decomposition of a model's latent space. This enables useful applications such as steering - in…

cs.CL2024

Function Vectors in Large Language Models

Eric Todd, Millicent L. Li, Arnab Sen Sharma +3

We report the presence of a simple neural mechanism that represents an input-output function as a vector within autoregressive transformer language models (LMs). Using causal media…

cs.CL2022

Label Semantic Aware Pre-training for Few-shot Text Classification

Aaron Mueller, Jason Krone, Salvatore Romeo +4

In text classification tasks, useful information is encoded in the label names. Label semantic aware systems have leveraged this information for improved text classification perfor…

cs.CL2022

What Do NLP Researchers Believe? Results of the NLP Community Metasurvey

Julian Michael, Ari Holtzman, Alicia Parrish +8

We present the results of the NLP Community Metasurvey. Run from May to June 2022, the survey elicited opinions on controversial issues, including industry influence in the field,…

cs.CL2020

Cross-Linguistic Syntactic Evaluation of Word Prediction Models

Aaron Mueller, Garrett Nicolai, Panayiota Petrou-Zeniou +2

A range of studies have concluded that neural word prediction models can distinguish grammatical from ungrammatical sentences with high accuracy. However, these studies are based p…

cs.CL2025

BabyLM Turns 3: Call for papers for the 2025 BabyLM workshop

Lucas Charpentier, Leshem Choshen, Ryan Cotterell +11

BabyLM aims to dissolve the boundaries between cognitive modeling and language modeling. We call for both workshop papers and for researchers to join the 3rd BabyLM competition. As…

cs.CL2023

Call for Papers -- The BabyLM Challenge: Sample-efficient pretraining on a developmentally plausible corpus

Alex Warstadt, Leshem Choshen, Aaron Mueller +3

We present the call for papers for the BabyLM Challenge: Sample-efficient pretraining on a developmentally plausible corpus. This shared task is intended for participants with an i…

cs.CL2025

Large Language Models Share Representations of Latent Grammatical Concepts Across Typologically Diverse Languages

Jannik Brinkmann, Chris Wendler, Christian Bartelt +1

Human bilinguals often use similar brain regions to process multiple languages, depending on when they learned their second language and their proficiency. In large language models…

cs.CL2026

CRISP: Persistent Concept Unlearning via Sparse Autoencoders

Tomer Ashuach, Dana Arad, Aaron Mueller +2

As large language models (LLMs) are increasingly deployed in real-world applications, the need to selectively remove unwanted knowledge while preserving model utility has become pa…

cs.LG2025

Sparse Feature Circuits: Discovering and Editing Interpretable Causal Graphs in Language Models

Samuel Marks, Can Rager, Eric J. Michaud +3

We introduce methods for discovering and applying sparse feature circuits. These are causally implicated subnetworks of human-interpretable features for explaining language model b…

cs.CL2022

Coloring the Blank Slate: Pre-training Imparts a Hierarchical Inductive Bias to Sequence-to-sequence Models

Aaron Mueller, Robert Frank, Tal Linzen +2

Relations between words are governed by hierarchical structure rather than linear ordering. Sequence-to-sequence (seq2seq) models, despite their success in downstream NLP applicati…

cs.CL2019

Quantity doesn't buy quality syntax with neural language models

Marten van Schijndel, Aaron Mueller, Tal Linzen

Recurrent neural networks can learn to predict upcoming words remarkably well on average; in syntactically complex contexts, however, they often assign unexpectedly high probabilit…

cs.LG2026

Two Stages of Folding: Convergent Mechanisms in AI Protein Folding Trunks

Kevin Lu, Jannik Brinkmann, Stefan Huber +4

How do protein structure prediction models fold proteins? We investigate this question through causal interventions on the folding trunks of ESMFold, OpenFold, and Boltz-1. Across…

cs.CL2025

Disjoint Processing Mechanisms of Hierarchical and Linear Grammars in Large Language Models

Aruna Sankaranarayanan, Dylan Hadfield-Menell, Aaron Mueller

All natural languages are structured hierarchically. In humans, this structural restriction is neurologically coded: when two grammars are presented with identical vocabularies, br…

cs.CL2025

Findings of the BlackboxNLP 2025 Shared Task: Localizing Circuits and Causal Variables in Language Models

Dana Arad, Yonatan Belinkov, Hanjie Chen +5

Mechanistic interpretability (MI) seeks to uncover how language models (LMs) implement specific behaviors, yet measuring progress in MI remains challenging. The recently released M…

cs.CL2025

Arithmetic Without Algorithms: Language Models Solve Math With a Bag of Heuristics

Yaniv Nikankin, Anja Reusch, Aaron Mueller +1

Do large language models (LLMs) solve reasoning tasks by learning robust generalizable algorithms, or do they memorize training data? To investigate this question, we use arithmeti…

cs.LG2025

NNsight and NDIF: Democratizing Access to Open-Weight Foundation Model Internals

Jaden Fiotto-Kaufman, Alexander R. Loftus, Eric Todd +17

We introduce NNsight and NDIF, technologies that work in tandem to enable scientific study of the representations and computations learned by very large neural networks. NNsight is…

cs.CL2021

Decoding Methods for Neural Narrative Generation

Alexandra DeLucia, Aaron Mueller, Xiang Lisa Li +1

Narrative generation is an open-ended NLP task in which a model generates a story given a prompt. The task is similar to neural response generation for chatbots; however, innovatio…

cs.CL2024

Findings of the Second BabyLM Challenge: Sample-Efficient Pretraining on Developmentally Plausible Corpora

Michael Y. Hu, Aaron Mueller, Candace Ross +7

The BabyLM Challenge is a community effort to close the data-efficiency gap between human and computational language learners. Participants compete to optimize language model train…

cs.CL2024

Inverse Scaling: When Bigger Isn't Better

Ian R. McKenzie, Alexander Lyzhov, Michael Pieler +24

Work on scaling laws has found that large language models (LMs) show predictable improvements to overall loss with increased scale (model size, training data, and compute). Here, w…

cs.CL2021

Causal Analysis of Syntactic Agreement Mechanisms in Neural Language Models

Matthew Finlayson, Aaron Mueller, Sebastian Gehrmann +3

Targeted syntactic evaluations have demonstrated the ability of language models to perform subject-verb agreement given difficult contexts. To elucidate the mechanisms by which the…

cs.CL2025

Measuring Mechanistic Independence: Can Bias Be Removed Without Erasing Demographics?

Zhengyang Shan, Aaron Mueller

We investigate how independent demographic bias mechanisms are from general demographic recognition in language models. Using a multi-task evaluation setup where demographics are a…

cs.LG2025

Priors in Time: Missing Inductive Biases for Language Model Interpretability

Ekdeep Singh Lubana, Can Rager, Sai Sumedh R. Hindupur +13

Recovering meaningful concepts from language model activations is a central aim of interpretability. While existing feature extraction methods aim to identify concepts that are ind…

cs.CL2024

Incremental Sentence Processing Mechanisms in Autoregressive Transformer Language Models

Michael Hanna, Aaron Mueller

Autoregressive transformer language models (LMs) possess strong syntactic abilities, often successfully handling phenomena from agreement to NPI licensing. However, the features th…

cs.CL2026

BabyLM Turns 4 and Goes Multilingual: Call for Papers for the 2026 BabyLM Workshop

Leshem Choshen, Ryan Cotterell, Mustafa Omer Gul +7

The goal of the BabyLM is to stimulate new research connections between cognitive modeling and language model pretraining. We invite contributions in this vein to the BabyLM Worksh…

cs.LG2024

Missed Causes and Ambiguous Effects: Counterfactuals Pose Challenges for Interpreting Neural Networks

Aaron Mueller

Interpretability research takes counterfactual theories of causality for granted. Most causal methods rely on counterfactual interventions to inputs or the activations of particula…

cs.CL2026

Do Language Models Track Entities Across State Changes?

Zilu Tang, Qiao Zhao, Gabriel Franco +4

Entity tracking (ET), the ability to keep track of states, is a fundamental skill that underlies complex reasoning. An increasing amount of work investigates how transformer langua…

cs.CL2021

Fine-tuning Encoders for Improved Monolingual and Zero-shot Polylingual Neural Topic Modeling

Aaron Mueller, Mark Dredze

Neural topic models can augment or replace bag-of-words inputs with the learned representations of deep pre-trained transformer-based word prediction models. One added benefit when…

cs.CL2026

Crosscoding Through Time: Tracking Emergence & Consolidation Of Linguistic Representations Throughout LLM Pretraining

Deniz Bayazit, Aaron Mueller, Antoine Bosselut

Large language models (LLMs) learn non-trivial abstractions during pretraining, such as detecting irregular plural noun subjects. However, because traditional evaluation methods (e…

cs.CL2025

Characterizing the Role of Similarity in the Property Inferences of Language Models

Juan Diego Rodriguez, Aaron Mueller, Kanishka Misra

Property inheritance -- a phenomenon where novel properties are projected from higher level categories (e.g., birds) to lower level ones (e.g., sparrows) -- provides a unique windo…

cs.LG2025

The Quest for the Right Mediator: Surveying Mechanistic Interpretability Through the Lens of Causal Mediation Analysis

Aaron Mueller, Jannik Brinkmann, Millicent Li +10

Interpretability provides a toolset for understanding how and why neural networks behave in certain ways. However, there is little unity in the field: most studies employ ad-hoc ev…

cs.LG2025

MIB: A Mechanistic Interpretability Benchmark

Aaron Mueller, Atticus Geiger, Sarah Wiegreffe +20

How can we know whether new mechanistic interpretability methods achieve real improvements? In pursuit of lasting evaluation standards, we propose MIB, a Mechanistic Interpretabili…

cs.CL2025

Findings of the BabyLM Challenge: Sample-Efficient Pretraining on Developmentally Plausible Corpora

Alex Warstadt, Aaron Mueller, Leshem Choshen +8

Children can acquire language from less than 100 million words of input. Large language models are far less data-efficient: they typically require 3 or 4 orders of magnitude more d…

cs.CL2019

Modeling Color Terminology Across Thousands of Languages

Arya D. McCarthy, Winston Wu, Aaron Mueller +2

There is an extensive history of scholarship into what constitutes a "basic" color term, as well as a broadly attested acquisition sequence of basic color terms across many languag…

cs.CV2026

BabyVLM-V2: Toward Developmentally Grounded Pretraining and Benchmarking of Vision Foundation Models

Shengao Wang, Wenqi Wang, Zecheng Wang +20

Early children's developmental trajectories set up a natural goal for sample-efficient pretraining of vision foundation models. We introduce BabyVLM-V2, a developmentally grounded…

cs.CL2022

Language model acceptability judgements are not always robust to context

Koustuv Sinha, Jon Gauthier, Aaron Mueller +4

Targeted syntactic evaluations of language models ask whether models show stable preferences for syntactically acceptable content over minimal-pair unacceptable inputs. Most target…

cs.CL2026

Improving the OOD Performance of Closed-Source LLMs on NLI Through Strategic Data Selection

Joe Stacey, Lisa Alazraki, Aran Ubhi +3

We investigate the robustness of fine-tuned Large Language Models (LLMs) for the task of Natural Language Inference (NLI), finding that the in-distribution gains from fine-tuning c…

cs.LG2025

Position-aware Automatic Circuit Discovery

Tal Haklay, Hadas Orgad, David Bau +2

A widely used strategy to discover and understand language model mechanisms is circuit analysis. A circuit is a minimal subgraph of a model's computation graph that executes a spec…

cs.CL2023

How to Plant Trees in Language Models: Data and Architectural Effects on the Emergence of Syntactic Inductive Biases

Aaron Mueller, Tal Linzen

Accurate syntactic representations are essential for robust generalization in natural language. Recent work has found that pre-training can teach language models to rely on hierarc…

cs.LG2026

The Curse of Multiple Mediators: Hidden Interaction Effects in Activation Patching

Sankaran Vaidyanathan, David Arbour, Aaron Mueller +2

Activation patching is the primary tool in mechanistic interpretability. It attributes causal responsibility for a model behavior to each of its individual components by estimating…

cs.AI2026

Pitfalls in Evaluating Interpretability Agents

Tal Haklay, Nikhil Prakash, Sana Pandey +5

Automated interpretability systems aim to reduce the need for human labor and scale analysis to increasingly large models and diverse tasks. Recent efforts toward this goal leverag…

cs.CL2026

In-Context Learning Without Copying

Kerem Sahin, Sheridan Feucht, Adam Belfki +4

Induction heads are attention heads that perform inductive copying by matching patterns from earlier context and copying their continuations verbatim. As models develop induction h…

cs.CL2024

In-context Learning Generalizes, But Not Always Robustly: The Case of Syntax

Aaron Mueller, Albert Webson, Jackson Petty +1

In-context learning (ICL) is now a common method for teaching large language models (LLMs) new tasks: given labeled examples in the input context, the LLM learns to perform the tas…

cs.CL2020

Demographic Representation and Collective Storytelling in the Me Too Twitter Hashtag Activism Movement

Aaron Mueller, Zach Wood-Doughty, Silvio Amir +2

The #MeToo movement on Twitter has drawn attention to the pervasive nature of sexual harassment and violence. While #MeToo has been praised for providing support for self-disclosur…

cs.AI2026

Latent Agents: A Post-Training Procedure for Internalized Multi-Agent Debate

John Seon Keun Yi, Aaron Mueller, Dokyun Lee

Multi-agent debate has been shown to improve reasoning in large language models (LLMs). However, it is compute-intensive, requiring generation of long transcripts before answering…

cs.CL2023

Meta-training with Demonstration Retrieval for Efficient Few-shot Learning

Aaron Mueller, Kanika Narang, Lambert Mathias +2

Large language models show impressive results on few-shot NLP tasks. However, these models are memory and computation-intensive. Meta-training allows one to leverage smaller models…

cs.LG2026

From Isolation to Entanglement: When Do Interpretability Methods Identify and Disentangle Known Concepts?

Aaron Mueller, Andrew Lee, Shruti Joshi +3

A goal of interpretability is to recover disentangled representations of latent concepts (features) from the activations of neural networks. The quality of features is typically ev…

cs.CL2024

[Call for Papers] The 2nd BabyLM Challenge: Sample-efficient pretraining on a developmentally plausible corpus

Leshem Choshen, Ryan Cotterell, Michael Y. Hu +7

After last year's successful BabyLM Challenge, the competition will be hosted again in 2024/2025. The overarching goals of the challenge remain the same; however, some of the compe…