@inproceedings{wang-etal-2026-action,
title = "Action Boundary Blindness: When {LLM} Agents Cannot Tell Where One Action Ends and Another Begins",
author = "Wang, Zhangyi and
Yu, Bingnan and
Xu, Jiexiang and
Li, Zongze",
editor = "Liakata, Maria and
Moreira, Viviane P. and
Zhang, Jiajun and
Jurgens, David",
booktitle = "Proceedings of the 64th Annual Meeting of the {A}ssociation for {C}omputational {L}inguistics (Volume 1: Long Papers)",
month = jul,
year = "2026",
address = "San Diego, California, United States",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.acl-long.1711/",
doi = "10.18653/v1/2026.acl-long.1711",
pages = "36883--36899",
ISBN = "979-8-89176-390-6",
abstract = "Large language model (LLM) agents excel at multi-step tasks yet frequently exhibit \textit{Action Boundary Blindness}{---}the inability to correctly determine action granularity, scope, and completeness. Grounded in Event Segmentation Theory from cognitive science, we formalize three violation types: \textit{granularity confusion}, \textit{scope creep}, and \textit{boundary ambiguity}. We propose four automatic metrics{---}Action Boundary Score (ABS), Granularity Alignment Rate (GAR), Scope Violation Rate (SVR), and Boundary-Aware Success Rate (BASR){---}requiring no human annotation. Experiments on 1,655 tasks across six benchmarks ($\tau$-bench, WebArena, ALFWorld, TheAgentCompany, OSWorld) with seven LLMs reveal that: (1) the best model achieves only 0.424 ABS; (2) using a multi-label attribution framework validated by inter-annotator agreement ($\kappa = 0.78$), boundary blindness is the primary failure mode in 37.2{\%} of failures (25.8{\%} as sole cause; 55.9{\%} total involvement including contributing factors); (3) under-action dominates at 48.4{\%}; (4) BASR is consistently $\sim$4 points lower than traditional success rate, exposing ``lucky successes.'' Critically, Explicit Boundary Prompting (EBP) improves ABS by 0.08{--}0.13 across all models, demonstrating that boundary blindness is better characterized as an \textit{elicitation gap} rather than a fundamental capability limitation{---}LLMs possess latent boundary perception not activated by default. This finding has implications for alignment and instruction tuning. We validate metrics through state-based cross-validation and human audit, estimating $\sim$22{\%} false positive rate from valid alternative paths, with model rankings remaining stable (Spearman $\rho = 1.0$)."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="wang-etal-2026-action">
<titleInfo>
<title>Action Boundary Blindness: When LLM Agents Cannot Tell Where One Action Ends and Another Begins</title>
</titleInfo>
<name type="personal">
<namePart type="given">Zhangyi</namePart>
<namePart type="family">Wang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Bingnan</namePart>
<namePart type="family">Yu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jiexiang</namePart>
<namePart type="family">Xu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Zongze</namePart>
<namePart type="family">Li</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2026-07</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)</title>
</titleInfo>
<name type="personal">
<namePart type="given">Maria</namePart>
<namePart type="family">Liakata</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Viviane</namePart>
<namePart type="given">P</namePart>
<namePart type="family">Moreira</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jiajun</namePart>
<namePart type="family">Zhang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">David</namePart>
<namePart type="family">Jurgens</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">San Diego, California, United States</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
<identifier type="isbn">979-8-89176-390-6</identifier>
</relatedItem>
<abstract>Large language model (LLM) agents excel at multi-step tasks yet frequently exhibit Action Boundary Blindness—the inability to correctly determine action granularity, scope, and completeness. Grounded in Event Segmentation Theory from cognitive science, we formalize three violation types: granularity confusion, scope creep, and boundary ambiguity. We propose four automatic metrics—Action Boundary Score (ABS), Granularity Alignment Rate (GAR), Scope Violation Rate (SVR), and Boundary-Aware Success Rate (BASR)—requiring no human annotation. Experiments on 1,655 tasks across six benchmarks (τ-bench, WebArena, ALFWorld, TheAgentCompany, OSWorld) with seven LLMs reveal that: (1) the best model achieves only 0.424 ABS; (2) using a multi-label attribution framework validated by inter-annotator agreement (ąppa = 0.78), boundary blindness is the primary failure mode in 37.2% of failures (25.8% as sole cause; 55.9% total involvement including contributing factors); (3) under-action dominates at 48.4%; (4) BASR is consistently \sim4 points lower than traditional success rate, exposing “lucky successes.” Critically, Explicit Boundary Prompting (EBP) improves ABS by 0.08–0.13 across all models, demonstrating that boundary blindness is better characterized as an elicitation gap rather than a fundamental capability limitation—LLMs possess latent boundary perception not activated by default. This finding has implications for alignment and instruction tuning. We validate metrics through state-based cross-validation and human audit, estimating \sim22% false positive rate from valid alternative paths, with model rankings remaining stable (Spearman ρ = 1.0).</abstract>
<identifier type="citekey">wang-etal-2026-action</identifier>
<identifier type="doi">10.18653/v1/2026.acl-long.1711</identifier>
<location>
<url>https://aclanthology.org/2026.acl-long.1711/</url>
</location>
<part>
<date>2026-07</date>
<extent unit="page">
<start>36883</start>
<end>36899</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Action Boundary Blindness: When LLM Agents Cannot Tell Where One Action Ends and Another Begins
%A Wang, Zhangyi
%A Yu, Bingnan
%A Xu, Jiexiang
%A Li, Zongze
%Y Liakata, Maria
%Y Moreira, Viviane P.
%Y Zhang, Jiajun
%Y Jurgens, David
%S Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)
%D 2026
%8 July
%I Association for Computational Linguistics
%C San Diego, California, United States
%@ 979-8-89176-390-6
%F wang-etal-2026-action
%X Large language model (LLM) agents excel at multi-step tasks yet frequently exhibit Action Boundary Blindness—the inability to correctly determine action granularity, scope, and completeness. Grounded in Event Segmentation Theory from cognitive science, we formalize three violation types: granularity confusion, scope creep, and boundary ambiguity. We propose four automatic metrics—Action Boundary Score (ABS), Granularity Alignment Rate (GAR), Scope Violation Rate (SVR), and Boundary-Aware Success Rate (BASR)—requiring no human annotation. Experiments on 1,655 tasks across six benchmarks (τ-bench, WebArena, ALFWorld, TheAgentCompany, OSWorld) with seven LLMs reveal that: (1) the best model achieves only 0.424 ABS; (2) using a multi-label attribution framework validated by inter-annotator agreement (ąppa = 0.78), boundary blindness is the primary failure mode in 37.2% of failures (25.8% as sole cause; 55.9% total involvement including contributing factors); (3) under-action dominates at 48.4%; (4) BASR is consistently \sim4 points lower than traditional success rate, exposing “lucky successes.” Critically, Explicit Boundary Prompting (EBP) improves ABS by 0.08–0.13 across all models, demonstrating that boundary blindness is better characterized as an elicitation gap rather than a fundamental capability limitation—LLMs possess latent boundary perception not activated by default. This finding has implications for alignment and instruction tuning. We validate metrics through state-based cross-validation and human audit, estimating \sim22% false positive rate from valid alternative paths, with model rankings remaining stable (Spearman ρ = 1.0).
%R 10.18653/v1/2026.acl-long.1711
%U https://aclanthology.org/2026.acl-long.1711/
%U https://doi.org/10.18653/v1/2026.acl-long.1711
%P 36883-36899
Markdown (Informal)
[Action Boundary Blindness: When LLM Agents Cannot Tell Where One Action Ends and Another Begins](https://aclanthology.org/2026.acl-long.1711/) (Wang et al., ACL 2026)
ACL