91 citations · 165 across the 33 of their papers we have counts for
35 papers
Why is AI-generated language so hard to understand?
Thomas Hikaru Clark, Matias B. Martin, Kyle Mahowald
There is a widespread sense that AI-generated text has recently become harder to understand. In this study, we investigate a set of psycholinguistic features that are known to corr…
Large dust grains escaping from a supernova remnant
Evgenii O. Vasiliev
Dust grains produced in the supernova (SN) ejecta at early ages are partially destroyed in hot gas of the SN remnant. Large grains owing to their inertia can overtake the forward s…
Causal Interventions Reveal Typologically Organized Syntactic Mechanisms in Multilingual Language Models
Sasha Boguraev, Toshiki Nakai, Kyle Mahowald +1
Linguistic theory has long recognized cross-linguistic syntactic regularities, leading to claims that these similar structures are processed by similar mechanisms. However, this hy…
VESTA: Visual Exploration with Statistical Tool Agents
William Rudman, Abhishek Divekar, Kanishk Jain +6
Fitting quantitative models to data is a central step in scientific workflows, yet it remains one of the least automated. Recent agent-based systems leverage language and vision-la…
The Counterexample Game: Iterated Conceptual Analysis and Repair in Language Models
Daniel Drucker, Kyle Mahowald
Conceptual analysis -- proposing definitions and refining them through counterexamples -- is central to philosophical methodology. We study whether language models can perform this…
Causal Drawbridges: Characterizing Gradient Blocking of Syntactic Islands in Transformer LMs
Sasha Boguraev, Kyle Mahowald
We show how causal interventions in Transformer models provide insights into English syntax by focusing on a long-standing challenge for syntactic theory: syntactic islands. Extrac…
Emergent Introspection in AI is Content-Agnostic
Harvey Lederman, Kyle Mahowald
Introspection is a foundational cognitive ability, but its mechanism is not well understood. Recent work has shown that AI models can introspect. We study the mechanism of this int…
When to Call an Apple Red: Humans Follow Introspective Rules, VLMs Don't
Jonathan Nemitz, Carsten Eickhoff, Junyi Jessy Li +3
Understanding when Vision-Language Models (VLMs) will behave unexpectedly, whether models can reliably predict their own behavior, and if models adhere to their introspective reaso…
From sunblock to softblock: Analyzing the correlates of neology in published writing and on social media
Maria Ryskina, Matthew R. Gormley, Kyle Mahowald +3
Living languages are shaped by a host of conflicting internal and external evolutionary pressures. While some of these pressures are universal across languages and cultures, others…
LLMs and people both learn to form conventions -- just not with each other
Cameron R. Jones, Agnese Lombardi, Kyle Mahowald +1
Humans align to one another in conversation -- adopting shared conventions that ease communication. We test whether LLMs form the same kinds of conventions in a multimodal communic…
France or Spain or Germany or France: A Neural Account of Non-Redundant Redundant Disjunctions
Sasha Boguraev, Qing Yao, Kyle Mahowald
Sentences like "She will go to France or Spain, or perhaps to Germany or France." appear formally redundant, yet become acceptable in contexts such as "Mary will go to a philosophy…
Mechanisms of Prompt-Induced Hallucination in Vision-Language Models
William Rudman, Michal Golovanevsky, Dana Arad +4
Large vision-language models (VLMs) are highly capable, yet often hallucinate by favoring textual prompts over visual evidence. We study this failure mode in a controlled object-co…
What Can String Probability Tell Us About Grammaticality?
Jennifer Hu, Ethan Gotlieb Wilcox, Siyuan Song +2
What have language models (LMs) learned about grammar? This question remains hotly debated, with major ramifications for linguistic theory. However, since probability and grammatic…
Convergence and Divergence of Language Models under Different Random Seeds
Finlay Fehlauer, Kyle Mahowald, Tiago Pimentel
In this paper, we investigate the convergence of language models (LMs) trained under different random seeds, measuring convergence as the expected per-token Kullback--Leibler (KL)…
Privileged Self-Access Matters for Introspection in AI
Siyuan Song, Harvey Lederman, Jennifer Hu +1
Whether AI models can introspect is an increasingly important practical question. But there is no consensus on how introspection is to be defined. Beginning from a recently propose…
semantic-features: A User-Friendly Tool for Studying Contextual Word Embeddings in Interpretable Semantic Spaces
Jwalanthi Ranganathan, Rohan Jha, Kanishka Misra +1
We introduce semantic-features, an extensible, easy-to-use library based on Chronis et al. (2023) for studying contextualized word embeddings of LMs by projecting them into interpr…
Is It JUST Semantics? A Case Study of Discourse Particle Understanding in LLMs
William Sheffield, Kanishka Misra, Valentina Pyatkin +3
Discourse particles are crucial elements that subtly shape the meaning of text. These words, often polyfunctional, give rise to nuanced and often quite disparate semantic/discourse…
A suite of LMs comprehend puzzle statements as well as humans
Adele E Goldberg, Supantho Rakshit, Jennifer Hu +1
Recent claims suggest that large language models (LMs) underperform humans in comprehending minimally complex English statements (Dentella et al., 2024). Here, we revisit those fin…
Causal Interventions Reveal Shared Structure Across English Filler-Gap Constructions
Sasha Boguraev, Christopher Potts, Kyle Mahowald
Language Models (LMs) have emerged as powerful sources of evidence for linguists seeking to develop theories of syntax. In this paper, we argue that causal interpretability methods…
On Language Models' Sensitivity to Suspicious Coincidences
Sriram Padmanabhan, Kanishka Misra, Kyle Mahowald +1
Humans are sensitive to suspicious coincidences when generalizing inductively over data, as they make assumptions as to how the data was sampled. This results in smaller, more spec…
Both Direct and Indirect Evidence Contribute to Dative Alternation Preferences in Language Models
Qing Yao, Kanishka Misra, Leonie Weissweiler +1
Language models (LMs) tend to show human-like preferences on a number of syntactic phenomena, but the extent to which these are attributable to direct exposure to the phenomena or…
Language Models Fail to Introspect About Their Knowledge of Language
Siyuan Song, Jennifer Hu, Kyle Mahowald
There has been recent interest in whether large language models (LLMs) can introspect about their own internal states. Such abilities would make LLMs more interpretable, and also v…
Constructions are Revealed in Word Distributions
Joshua Rozner, Leonie Weissweiler, Kyle Mahowald +1
Construction grammar posits that constructions, or form-meaning pairings, are acquired through experience with language (the distributional learning hypothesis). But how much infor…
Linguistic Generalizations are not Rules: Impacts on Evaluation of LMs
Leonie Weissweiler, Kyle Mahowald, Adele Goldberg
Linguistic evaluations of how well LMs generalize to produce or understand language often implicitly take for granted that natural languages are generated by symbolic rules. Accord…