<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="xu-etal-2026-failalign">
    <titleInfo>
        <title>FailAlign: Failure-Aligned Bootstrapping for Long-Horizon HTML Based Code Generation Agents</title>
    </titleInfo>
    <name type="personal">
        <namePart type="given">Fanglin</namePart>
        <namePart type="family">Xu</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Wei</namePart>
        <namePart type="family">Zhang</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Jian</namePart>
        <namePart type="family">Yang</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Guo</namePart>
        <namePart type="family">Chen</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Liqun</namePart>
        <namePart type="family">Yang</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Zhoujun</namePart>
        <namePart type="family">Li</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>LLM-based web agents automate browser interactions by generating structured action sequences, yet their deployment in long-horizon tasks is hindered by fragile error propagation in noisy DOM trees. Critically, existing training paradigms rely exclusively on positive supervision from gold demonstrations, teaching models **what** is correct but never **why** plausible alternatives fail, leaving agents unable to distinguish between subtly different yet critically divergent choices. We propose **FailAlign**, a failure-aligned learning framework that systematically integrates error signals as first-class training supervision. We mine a Failure Index from model rollouts by pinpointing first-error steps, then leverage these instances across a three-stage pipeline: (i) action-grounded context compression distills verbose HTML into compact DOM snippets; (ii) hard-negative DPO pairs gold actions with semantically similar incorrect alternatives to sharpen decision boundaries; (iii) trajectory-level GRPO mitigates error cascades by re-simulating corrected trajectories. On Mind2Web with Qwen3-32B, FailAlign achieves **49.7% SSR** and **5.6% SR**, surpassing models 15× larger (Qwen3-Coder-480B: 37.0% SSR) and commercial systems (Claude Opus-4.5: 39.4% SSR), validating the efficacy of learning from failures.</abstract>
    <identifier type="citekey">xu-etal-2026-failalign</identifier>
    <location>
        <url>https://aclanthology.org/2026.findings-emnlp.724/</url>
    </location>
    <part>
        <date>2026-10</date>
        <extent unit="page">
            <start>15223</start>
            <end>15236</end>
        </extent>
    </part>
</mods>
</modsCollection>

%0 Conference Proceedings
%T FailAlign: Failure-Aligned Bootstrapping for Long-Horizon HTML Based Code Generation Agents
%A Xu, Fanglin
%A Zhang, Wei
%A Yang, Jian
%A Chen, Guo
%A Yang, Liqun
%A Li, Zhoujun
%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 xu-etal-2026-failalign
%X LLM-based web agents automate browser interactions by generating structured action sequences, yet their deployment in long-horizon tasks is hindered by fragile error propagation in noisy DOM trees. Critically, existing training paradigms rely exclusively on positive supervision from gold demonstrations, teaching models **what** is correct but never **why** plausible alternatives fail, leaving agents unable to distinguish between subtly different yet critically divergent choices. We propose **FailAlign**, a failure-aligned learning framework that systematically integrates error signals as first-class training supervision. We mine a Failure Index from model rollouts by pinpointing first-error steps, then leverage these instances across a three-stage pipeline: (i) action-grounded context compression distills verbose HTML into compact DOM snippets; (ii) hard-negative DPO pairs gold actions with semantically similar incorrect alternatives to sharpen decision boundaries; (iii) trajectory-level GRPO mitigates error cascades by re-simulating corrected trajectories. On Mind2Web with Qwen3-32B, FailAlign achieves **49.7% SSR** and **5.6% SR**, surpassing models 15× larger (Qwen3-Coder-480B: 37.0% SSR) and commercial systems (Claude Opus-4.5: 39.4% SSR), validating the efficacy of learning from failures.
%U https://aclanthology.org/2026.findings-emnlp.724/
%P 15223-15236

Markdown (Informal)

[FailAlign: Failure-Aligned Bootstrapping for Long-Horizon HTML Based Code Generation Agents](https://aclanthology.org/2026.findings-emnlp.724/) (Xu et al., Findings 2026)

• FailAlign: Failure-Aligned Bootstrapping for Long-Horizon HTML Based Code Generation Agents (Xu et al., Findings 2026)

ACL

• Fanglin Xu, Wei Zhang, Jian Yang, Guo Chen, Liqun Yang, and Zhoujun Li. 2026. FailAlign: Failure-Aligned Bootstrapping for Long-Horizon HTML Based Code Generation Agents. In Findings of the Association for Computational Linguistics: EMNLP 2026, pages 15223–15236, Budapest, Hungary. Association for Computational Linguistics.