@inproceedings{lepori-etal-2025-racing,
title = "Racing Thoughts: Explaining Contextualization Errors in Large Language Models",
author = "Lepori, Michael A. and
Mozer, Michael Curtis and
Ghandeharioun, Asma",
editor = "Chiruzzo, Luis and
Ritter, Alan and
Wang, Lu",
booktitle = "Proceedings of the 2025 Conference of the Nations of the Americas Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 1: Long Papers)",
month = apr,
year = "2025",
address = "Albuquerque, New Mexico",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2025.naacl-long.155/",
doi = "10.18653/v1/2025.naacl-long.155",
pages = "3020--3036",
ISBN = "979-8-89176-189-6",
abstract = "The profound success of transformer-based language models can largely be attributed to their ability to integrate relevant contextual information from an input sequence in order to generate a response or complete a task. However, we know very little about the algorithms that a model employs to implement this capability, nor do we understand their failure modes. For example, given the prompt ``John is going fishing, so he walks over to the bank. Can he make an ATM transaction?'', a model may incorrectly respond ``Yes'' if it has not properly contextualized ``bank'' as a geographical feature, rather than a financial institution. We propose the LLM Race Conditions Hypothesis as an explanation of contextualization errors of this form. This hypothesis identifies dependencies between tokens (e.g., ``bank'' must be properly contextualized before the final token, ``?'', integrates information from ``bank''), and claims that contextualization errors are a result of violating these dependencies. Using a variety of techniques from mechanistic interpretability, we provide correlational and causal evidence in support of the hypothesis and suggest inference-time interventions to address it."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="lepori-etal-2025-racing">
<titleInfo>
<title>Racing Thoughts: Explaining Contextualization Errors in Large Language Models</title>
</titleInfo>
<name type="personal">
<namePart type="given">Michael</namePart>
<namePart type="given">A</namePart>
<namePart type="family">Lepori</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Michael</namePart>
<namePart type="given">Curtis</namePart>
<namePart type="family">Mozer</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Asma</namePart>
<namePart type="family">Ghandeharioun</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2025-04</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Proceedings of the 2025 Conference of the Nations of the Americas Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 1: Long Papers)</title>
</titleInfo>
<name type="personal">
<namePart type="given">Luis</namePart>
<namePart type="family">Chiruzzo</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Alan</namePart>
<namePart type="family">Ritter</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Lu</namePart>
<namePart type="family">Wang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Albuquerque, New Mexico</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
<identifier type="isbn">979-8-89176-189-6</identifier>
</relatedItem>
<abstract>The profound success of transformer-based language models can largely be attributed to their ability to integrate relevant contextual information from an input sequence in order to generate a response or complete a task. However, we know very little about the algorithms that a model employs to implement this capability, nor do we understand their failure modes. For example, given the prompt “John is going fishing, so he walks over to the bank. Can he make an ATM transaction?”, a model may incorrectly respond “Yes” if it has not properly contextualized “bank” as a geographical feature, rather than a financial institution. We propose the LLM Race Conditions Hypothesis as an explanation of contextualization errors of this form. This hypothesis identifies dependencies between tokens (e.g., “bank” must be properly contextualized before the final token, “?”, integrates information from “bank”), and claims that contextualization errors are a result of violating these dependencies. Using a variety of techniques from mechanistic interpretability, we provide correlational and causal evidence in support of the hypothesis and suggest inference-time interventions to address it.</abstract>
<identifier type="citekey">lepori-etal-2025-racing</identifier>
<identifier type="doi">10.18653/v1/2025.naacl-long.155</identifier>
<location>
<url>https://aclanthology.org/2025.naacl-long.155/</url>
</location>
<part>
<date>2025-04</date>
<extent unit="page">
<start>3020</start>
<end>3036</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Racing Thoughts: Explaining Contextualization Errors in Large Language Models
%A Lepori, Michael A.
%A Mozer, Michael Curtis
%A Ghandeharioun, Asma
%Y Chiruzzo, Luis
%Y Ritter, Alan
%Y Wang, Lu
%S Proceedings of the 2025 Conference of the Nations of the Americas Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 1: Long Papers)
%D 2025
%8 April
%I Association for Computational Linguistics
%C Albuquerque, New Mexico
%@ 979-8-89176-189-6
%F lepori-etal-2025-racing
%X The profound success of transformer-based language models can largely be attributed to their ability to integrate relevant contextual information from an input sequence in order to generate a response or complete a task. However, we know very little about the algorithms that a model employs to implement this capability, nor do we understand their failure modes. For example, given the prompt “John is going fishing, so he walks over to the bank. Can he make an ATM transaction?”, a model may incorrectly respond “Yes” if it has not properly contextualized “bank” as a geographical feature, rather than a financial institution. We propose the LLM Race Conditions Hypothesis as an explanation of contextualization errors of this form. This hypothesis identifies dependencies between tokens (e.g., “bank” must be properly contextualized before the final token, “?”, integrates information from “bank”), and claims that contextualization errors are a result of violating these dependencies. Using a variety of techniques from mechanistic interpretability, we provide correlational and causal evidence in support of the hypothesis and suggest inference-time interventions to address it.
%R 10.18653/v1/2025.naacl-long.155
%U https://aclanthology.org/2025.naacl-long.155/
%U https://doi.org/10.18653/v1/2025.naacl-long.155
%P 3020-3036
Markdown (Informal)
[Racing Thoughts: Explaining Contextualization Errors in Large Language Models](https://aclanthology.org/2025.naacl-long.155/) (Lepori et al., NAACL 2025)
ACL
- Michael A. Lepori, Michael Curtis Mozer, and Asma Ghandeharioun. 2025. Racing Thoughts: Explaining Contextualization Errors in Large Language Models. In Proceedings of the 2025 Conference of the Nations of the Americas Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 1: Long Papers), pages 3020–3036, Albuquerque, New Mexico. Association for Computational Linguistics.