<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="liu-etal-2026-adaptiveguard">
    <titleInfo>
        <title>AdaptiveGuard: Treating Constraints as Falsifiable Hypotheses for LLM Agents in Dynamic Interactive Environments</title>
    </titleInfo>
    <name type="personal">
        <namePart type="given">Pengxiang</namePart>
        <namePart type="family">Liu</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Tao</namePart>
        <namePart type="family">Ren</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Wei</namePart>
        <namePart type="family">Xiong</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Lijie</namePart>
        <namePart type="family">Li</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Junjie</namePart>
        <namePart type="family">Wang</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Jun</namePart>
        <namePart type="family">Hu</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>Findings of the Association for Computational Linguistics: EMNLP 2026</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">979-8-89176-462-0</identifier>
    </relatedItem>
    <abstract>Large Language Models (LLMs) have shown impressive reasoning capabilities in agents for complex interactive environments. However, these agents frequently propose actions that violate physical or procedural constraints. Recent neuro-symbolic guards mitigate such errors by compiling explicit admissibility rules from oracle demonstrations, yet treat them as fixed laws—a closed-world assumption that breaks down under sparse demonstrations or environment shift, leaving either missing rules that fail to block unsafe actions or stale rules that wrongly reject valid alternatives. To address this, we introduce **AdaptiveGuard** ([code](https://github.com/lpxwork666111/AdaptiveGuard)), an Adaptive framework that Guards LLM agents by treating constraints as *falsifiable hypotheses*. Each rule is represented by a Beta-distributed validity belief mapped to *tentative*, *confirmed*, and *deprecated* tiers; uncertain rule blockers are then challenged through a zero-extra-call falsification gate; and these rule beliefs are continuously updated via contrastive feedback attribution from environment evidence. Benchmark experiments on ScienceWorld and VirtualHome show aggregate gains over strong static constraint guards, especially in novel scenarios. Complementary post-hoc diagnostics under sparse supervision and controlled shift show improved adaptation, while observed failed-trial runs remain below the configured budget cap and the conservative-bias proxy declines over the measured horizon.</abstract>
    <identifier type="citekey">liu-etal-2026-adaptiveguard</identifier>
    <location>
        <url>https://aclanthology.org/2026.findings-emnlp.1193/</url>
    </location>
    <part>
        <date>2026-10</date>
        <extent unit="page">
            <start>25186</start>
            <end>25205</end>
        </extent>
    </part>
</mods>
</modsCollection>

%0 Conference Proceedings
%T AdaptiveGuard: Treating Constraints as Falsifiable Hypotheses for LLM Agents in Dynamic Interactive Environments
%A Liu, Pengxiang
%A Ren, Tao
%A Xiong, Wei
%A Li, Lijie
%A Wang, Junjie
%A Hu, Jun
%Y Lee, Hung-Yi
%Y Dev, Sunipa
%Y Li, Jessy
%Y Elliott, Desmond
%S Findings of the Association for Computational Linguistics: EMNLP 2026
%D 2026
%8 October
%I Association for Computational Linguistics
%C Budapest, Hungary
%@ 979-8-89176-462-0
%F liu-etal-2026-adaptiveguard
%X Large Language Models (LLMs) have shown impressive reasoning capabilities in agents for complex interactive environments. However, these agents frequently propose actions that violate physical or procedural constraints. Recent neuro-symbolic guards mitigate such errors by compiling explicit admissibility rules from oracle demonstrations, yet treat them as fixed laws—a closed-world assumption that breaks down under sparse demonstrations or environment shift, leaving either missing rules that fail to block unsafe actions or stale rules that wrongly reject valid alternatives. To address this, we introduce **AdaptiveGuard** ([code](https://github.com/lpxwork666111/AdaptiveGuard)), an Adaptive framework that Guards LLM agents by treating constraints as *falsifiable hypotheses*. Each rule is represented by a Beta-distributed validity belief mapped to *tentative*, *confirmed*, and *deprecated* tiers; uncertain rule blockers are then challenged through a zero-extra-call falsification gate; and these rule beliefs are continuously updated via contrastive feedback attribution from environment evidence. Benchmark experiments on ScienceWorld and VirtualHome show aggregate gains over strong static constraint guards, especially in novel scenarios. Complementary post-hoc diagnostics under sparse supervision and controlled shift show improved adaptation, while observed failed-trial runs remain below the configured budget cap and the conservative-bias proxy declines over the measured horizon.
%U https://aclanthology.org/2026.findings-emnlp.1193/
%P 25186-25205

Markdown (Informal)

[AdaptiveGuard: Treating Constraints as Falsifiable Hypotheses for LLM Agents in Dynamic Interactive Environments](https://aclanthology.org/2026.findings-emnlp.1193/) (Liu et al., Findings 2026)

• AdaptiveGuard: Treating Constraints as Falsifiable Hypotheses for LLM Agents in Dynamic Interactive Environments (Liu et al., Findings 2026)

ACL

• Pengxiang Liu, Tao Ren, Wei Xiong, Lijie Li, Junjie Wang, and Jun Hu. 2026. AdaptiveGuard: Treating Constraints as Falsifiable Hypotheses for LLM Agents in Dynamic Interactive Environments. In Findings of the Association for Computational Linguistics: EMNLP 2026, pages 25186–25205, Budapest, Hungary. Association for Computational Linguistics.