Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

LLM Agents in Law: Taxonomy, Applications, and Challenges

Shuang Liu, Ruijia Zhang, Ruoyun Ma, Yujia Deng, Lanyi Zhu, Jiayu Li, Zelong Li, Zhibin Shen, Mengnan Du

TL;DR

The paper surveys the rise of LLM agents in the legal domain, addressing how agentic architectures—featuring planning, memory, tool usage, and grounding—overcome persistent issues of standalone LLMs such as hallucination, outdated information, and verifiability. It provides a structured taxonomy of five legal practice areas, analyzes representative academic and commercial systems, and reviews evaluation methodologies tailored to agentic law tasks. The authors identify open challenges (long-horizon reliability, citation fidelity, multi-agent coordination, jurisdictional diversity) and offer future directions emphasizing human-in-the-loop oversight, new law-specific benchmarks, and governance frameworks. The work aims to guide the development of robust, autonomous legal assistants with practical impact across research and industry practice.

Abstract

Large language models (LLMs) have precipitated a dramatic improvement in the legal domain, yet the deployment of standalone models faces significant limitations regarding hallucination, outdated information, and verifiability. Recently, LLM agents have attracted significant attention as a solution to these challenges, utilizing advanced capabilities such as planning, memory, and tool usage to meet the rigorous standards of legal practice. In this paper, we present a comprehensive survey of LLM agents for legal tasks, analyzing how these architectures bridge the gap between technical capabilities and domain-specific needs. Our major contributions include: (1) systematically analyzing the technical transition from standard legal LLMs to legal agents; (2) presenting a structured taxonomy of current agent applications across distinct legal practice areas; (3) discussing evaluation methodologies specifically for agentic performance in law; and (4) identifying open challenges and outlining future directions for developing robust and autonomous legal assistants.

LLM Agents in Law: Taxonomy, Applications, and Challenges

TL;DR

The paper surveys the rise of LLM agents in the legal domain, addressing how agentic architectures—featuring planning, memory, tool usage, and grounding—overcome persistent issues of standalone LLMs such as hallucination, outdated information, and verifiability. It provides a structured taxonomy of five legal practice areas, analyzes representative academic and commercial systems, and reviews evaluation methodologies tailored to agentic law tasks. The authors identify open challenges (long-horizon reliability, citation fidelity, multi-agent coordination, jurisdictional diversity) and offer future directions emphasizing human-in-the-loop oversight, new law-specific benchmarks, and governance frameworks. The work aims to guide the development of robust, autonomous legal assistants with practical impact across research and industry practice.

Abstract

Large language models (LLMs) have precipitated a dramatic improvement in the legal domain, yet the deployment of standalone models faces significant limitations regarding hallucination, outdated information, and verifiability. Recently, LLM agents have attracted significant attention as a solution to these challenges, utilizing advanced capabilities such as planning, memory, and tool usage to meet the rigorous standards of legal practice. In this paper, we present a comprehensive survey of LLM agents for legal tasks, analyzing how these architectures bridge the gap between technical capabilities and domain-specific needs. Our major contributions include: (1) systematically analyzing the technical transition from standard legal LLMs to legal agents; (2) presenting a structured taxonomy of current agent applications across distinct legal practice areas; (3) discussing evaluation methodologies specifically for agentic performance in law; and (4) identifying open challenges and outlining future directions for developing robust and autonomous legal assistants.
Paper Structure (33 sections, 3 figures)

This paper contains 33 sections, 3 figures.

Table of Contents

  1. Introduction
  2. Contribution and Uniqueness
  3. Why LLM Agents for Legal Tasks?
  4. Why LLMs by themselves are insufficient
  5. LLM Agents Address These Limitations
  6. Current LLM Agent Application Areas in the Legal Domain
  7. Legal Search & Research
  8. Legal Retrieval.
  9. Legal Theory Understanding & Legal Reasoning.
  10. Litigation & Dispute Resolution
  11. Judgment Prediction.
  12. Courtroom & Dispute Resolution Simulation.
  13. Compliance, Governance, & Regulation
  14. Compliance.
  15. Risk Assessment & Fraud detection.
  16. ...and 18 more sections

Figures (3)

  • Figure 1: Overview of paper structure and LLM agent application.
  • Figure 2: Taxonomy of LLM agents for legal tasks. This framework categorizes five core legal domains and maps them to representative academic and commercial agentic systems.
  • Figure 5: Legal Benchmarks Summary