@inproceedings{zuo-etal-2026-legalworld,
title = "{L}egal{W}orld: A Life-Cycle Interactive Environment for Legal Agents",
author = "Zuo, Songhan and
Yue, Shengbin and
Chiang, Tao and
LI, Guanying and
Song, Yun and
Huang, Xuanjing and
Wei, Zhongyu",
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.390/",
pages = "8942--8976",
ISBN = "979-8-89176-459-0",
abstract = "Civil litigation is inherently a life-cycle process: what a lawyer drafts on day one constrains what unfolds at trial months later. Yet existing legal benchmarks evaluate isolated subtasks, and prior legal-agent simulators reinitialize each scenario from shared ground truth, leaving cross-stage causal dependencies unmodeled. We present LegalWorld, a life-cycle interactive environment that models Chinese civil litigation as a causally connected state chain of five stages (seven sub-scenarios), grounded in 75,309 paired Chinese civil judgments. We pair it with reusable infrastructure (local memory, global case memory, a Skill/Tool library) and Reflective Legal Skill (RLS), which distills completed trajectories into reusable procedural skills. Building on this environment, we construct LongJud-Bench to evaluate agent capability across all five connected stages. 18,992 ratings from 217 legal-background evaluators confirm that LegalWorld trajectories are procedurally faithful and role-consistent; cross-model evaluation reveals stage-level divergences invisible to aggregate scores (backbones within 4 overall points can differ by up to 15 normalized points on individual stages), and RLS yields consistent gains on both reflected and held-out cases. Detailed resources will be provided upon acceptance."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="zuo-etal-2026-legalworld">
<titleInfo>
<title>LegalWorld: A Life-Cycle Interactive Environment for Legal Agents</title>
</titleInfo>
<name type="personal">
<namePart type="given">Songhan</namePart>
<namePart type="family">Zuo</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Shengbin</namePart>
<namePart type="family">Yue</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Tao</namePart>
<namePart type="family">Chiang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Guanying</namePart>
<namePart type="family">LI</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yun</namePart>
<namePart type="family">Song</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Xuanjing</namePart>
<namePart type="family">Huang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Zhongyu</namePart>
<namePart type="family">Wei</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>Civil litigation is inherently a life-cycle process: what a lawyer drafts on day one constrains what unfolds at trial months later. Yet existing legal benchmarks evaluate isolated subtasks, and prior legal-agent simulators reinitialize each scenario from shared ground truth, leaving cross-stage causal dependencies unmodeled. We present LegalWorld, a life-cycle interactive environment that models Chinese civil litigation as a causally connected state chain of five stages (seven sub-scenarios), grounded in 75,309 paired Chinese civil judgments. We pair it with reusable infrastructure (local memory, global case memory, a Skill/Tool library) and Reflective Legal Skill (RLS), which distills completed trajectories into reusable procedural skills. Building on this environment, we construct LongJud-Bench to evaluate agent capability across all five connected stages. 18,992 ratings from 217 legal-background evaluators confirm that LegalWorld trajectories are procedurally faithful and role-consistent; cross-model evaluation reveals stage-level divergences invisible to aggregate scores (backbones within 4 overall points can differ by up to 15 normalized points on individual stages), and RLS yields consistent gains on both reflected and held-out cases. Detailed resources will be provided upon acceptance.</abstract>
<identifier type="citekey">zuo-etal-2026-legalworld</identifier>
<location>
<url>https://aclanthology.org/2026.emnlp-main.390/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>8942</start>
<end>8976</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T LegalWorld: A Life-Cycle Interactive Environment for Legal Agents
%A Zuo, Songhan
%A Yue, Shengbin
%A Chiang, Tao
%A LI, Guanying
%A Song, Yun
%A Huang, Xuanjing
%A Wei, Zhongyu
%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 zuo-etal-2026-legalworld
%X Civil litigation is inherently a life-cycle process: what a lawyer drafts on day one constrains what unfolds at trial months later. Yet existing legal benchmarks evaluate isolated subtasks, and prior legal-agent simulators reinitialize each scenario from shared ground truth, leaving cross-stage causal dependencies unmodeled. We present LegalWorld, a life-cycle interactive environment that models Chinese civil litigation as a causally connected state chain of five stages (seven sub-scenarios), grounded in 75,309 paired Chinese civil judgments. We pair it with reusable infrastructure (local memory, global case memory, a Skill/Tool library) and Reflective Legal Skill (RLS), which distills completed trajectories into reusable procedural skills. Building on this environment, we construct LongJud-Bench to evaluate agent capability across all five connected stages. 18,992 ratings from 217 legal-background evaluators confirm that LegalWorld trajectories are procedurally faithful and role-consistent; cross-model evaluation reveals stage-level divergences invisible to aggregate scores (backbones within 4 overall points can differ by up to 15 normalized points on individual stages), and RLS yields consistent gains on both reflected and held-out cases. Detailed resources will be provided upon acceptance.
%U https://aclanthology.org/2026.emnlp-main.390/
%P 8942-8976
Markdown (Informal)
[LegalWorld: A Life-Cycle Interactive Environment for Legal Agents](https://aclanthology.org/2026.emnlp-main.390/) (Zuo et al., EMNLP 2026)
ACL
- Songhan Zuo, Shengbin Yue, Tao Chiang, Guanying LI, Yun Song, Xuanjing Huang, and Zhongyu Wei. 2026. LegalWorld: A Life-Cycle Interactive Environment for Legal Agents. In Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing, pages 8942–8976, Budapest, Hungary. Association for Computational Linguistics.