@inproceedings{cao-schuler-2026-norms,
title = "Norms of Error Terms in Transformer Language Models Predict Human Reading Time Beyond Surprisal",
author = "Cao, Ziyuan and
Schuler, William",
editor = "Lee, Hung-Yi and
Dev, Sunipa and
Li, Jessy and
Elliott, Desmond",
booktitle = "Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing",
month = oct,
year = "2026",
address = "Budapest, Hungary",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.emnlp-main.2364/",
doi = "10.18653/2026.emnlp-main.2364",
pages = "53691--53699",
ISBN = "979-8-89176-459-0",
abstract = "It is possible that expectation-based latency in reading time is caused by mechanistic latency in changing internal activation states to match observations; this is a primary motivation for predictive coding models (Ohams et al., 2026). This paper explores measurement of sentence processing effort that is based on the gradient signals of transformer language models during error backpropagation: the 2-norm of the gradient of surprisal w.r.t. the output of intermediate layers, i.e., the norm of the error terms at intermediate layers. Regression experiments on multiple human reading time datasets show that the norm of error terms at the output of intermediate layers of GPT-2 small add predictive power beyond baseline predictors including standard surprisal from the same model."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="cao-schuler-2026-norms">
<titleInfo>
<title>Norms of Error Terms in Transformer Language Models Predict Human Reading Time Beyond Surprisal</title>
</titleInfo>
<name type="personal">
<namePart type="given">Ziyuan</namePart>
<namePart type="family">Cao</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">William</namePart>
<namePart type="family">Schuler</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2026-10</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing</title>
</titleInfo>
<name type="personal">
<namePart type="given">Hung-Yi</namePart>
<namePart type="family">Lee</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Sunipa</namePart>
<namePart type="family">Dev</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jessy</namePart>
<namePart type="family">Li</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Desmond</namePart>
<namePart type="family">Elliott</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Budapest, Hungary</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
<identifier type="isbn">979-8-89176-459-0</identifier>
</relatedItem>
<abstract>It is possible that expectation-based latency in reading time is caused by mechanistic latency in changing internal activation states to match observations; this is a primary motivation for predictive coding models (Ohams et al., 2026). This paper explores measurement of sentence processing effort that is based on the gradient signals of transformer language models during error backpropagation: the 2-norm of the gradient of surprisal w.r.t. the output of intermediate layers, i.e., the norm of the error terms at intermediate layers. Regression experiments on multiple human reading time datasets show that the norm of error terms at the output of intermediate layers of GPT-2 small add predictive power beyond baseline predictors including standard surprisal from the same model.</abstract>
<identifier type="citekey">cao-schuler-2026-norms</identifier>
<identifier type="doi">10.18653/2026.emnlp-main.2364</identifier>
<location>
<url>https://aclanthology.org/2026.emnlp-main.2364/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>53691</start>
<end>53699</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Norms of Error Terms in Transformer Language Models Predict Human Reading Time Beyond Surprisal
%A Cao, Ziyuan
%A Schuler, William
%Y Lee, Hung-Yi
%Y Dev, Sunipa
%Y Li, Jessy
%Y Elliott, Desmond
%S Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing
%D 2026
%8 October
%I Association for Computational Linguistics
%C Budapest, Hungary
%@ 979-8-89176-459-0
%F cao-schuler-2026-norms
%X It is possible that expectation-based latency in reading time is caused by mechanistic latency in changing internal activation states to match observations; this is a primary motivation for predictive coding models (Ohams et al., 2026). This paper explores measurement of sentence processing effort that is based on the gradient signals of transformer language models during error backpropagation: the 2-norm of the gradient of surprisal w.r.t. the output of intermediate layers, i.e., the norm of the error terms at intermediate layers. Regression experiments on multiple human reading time datasets show that the norm of error terms at the output of intermediate layers of GPT-2 small add predictive power beyond baseline predictors including standard surprisal from the same model.
%R 10.18653/2026.emnlp-main.2364
%U https://aclanthology.org/2026.emnlp-main.2364/
%U https://doi.org/10.18653/2026.emnlp-main.2364
%P 53691-53699
Markdown (Informal)
[Norms of Error Terms in Transformer Language Models Predict Human Reading Time Beyond Surprisal](https://aclanthology.org/2026.emnlp-main.2364/) (Cao & Schuler, EMNLP 2026)
ACL