arXiv:2606.18728cs.CL2026-06

构建中国民事诉讼全周期交互环境,模拟律师行为的长期因果影响。

LegalWorld: A Life-Cycle Interactive Environment for Legal Agents

论文配图:LegalWorld: A Life-Cycle Interactive Environment for Legal Agents
图 1 · 摘自论文原文
  • 将民事诉讼建模为五个阶段的因果状态链,保持全程一致性。
  • 基于7.5万份判决数据,验证轨迹程序合规且角色一致。
  • 揭示单一模型无法在咨询、起草、庭审中全面领先,适合评估法律智能体综合能力。

民事诉讼本质上是一个生命周期过程:律师首日的文书工作会限制数月后庭审的展开。然而现有法律评测基准仅评估孤立子任务,以往的法律代理仿真器每次重置场景时依赖共享真实情况,未建模跨阶段因果依赖。我们提出LegalWorld,一个以中国民事诉讼为蓝本的生命周期交互环境,将其建模为五个阶段(七个子场景)的因果状态链,基于75,309对中文民事判决。配套可复用基础设施(本地记忆、全局案件记忆、技能/工具库)确保每桩纠纷在完整生命周期中保持一致。在此基础上,我们构建LongJud-Bench,评估代理在所有五个连接阶段的能力。217名具有法律背景的评估者给出18,992条评分,证实LegalWorld轨迹在程序上忠实且角色一致;跨模型能力评估显示显著差异,聚合分数无法揭示这些差异,无单一主干模型能在咨询、起草与法庭辩护中全面领先。详细资源将公开发布。

原文摘要 · Abstract (English)

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) that keeps each dispute consistent across its full life cycle. 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; and a capability-level cross-model evaluation reveals sharp divergences that aggregate scores cannot expose, with no single backbone leading across consultation, drafting, and courtroom advocacy. Detailed resources will be released publicly.

法律智能体生命周期建模多阶段评估中文司法

Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品;中文卡片由大模型生成,请以原文为准。