2024 LAWLegalAgenticWorkflowsforCust
- (Watson et al., 2024) ⇒ William Watson, Nicole Cho, Nishan Srishankar, Zhen Zeng, Lucas Cecchi, Daniel Scott, Suchetha Siddagangappa, Rachneet Kaur, Tucker Balch, and Manuela Veloso. (2024). “LAW: Legal Agentic Workflows for Custody and Fund Services Contracts.” In: arXiv preprint arXiv:2412.11063.
Subject Headings: Legal Domain-Specific NLP, Multi-Hop Reasoning, Legal Knowledge Representation.
Notes
- The paper introduces LAW, a modular AI framework designed to handle custody and fund services contracts by leveraging domain-specific tools and text agents.
- The paper highlights the limitations of fine-tuned legal LLMs, including context window constraints, hallucinations, and high computational costs, and presents LAW as a cost-effective alternative.
- The paper describes the integration of tools for direct extraction (e.g., extracting dates and parties) and multi-hop reasoning (e.g., calculating termination dates) within the LAW framework.
- The paper showcases LAW's ability to outperform GPT-3.5-turbo in retrieval and analytical tasks, achieving up to 92.9% accuracy gains in complex legal queries.
- The paper employs data from the SEC’s EDGAR database, specifically focusing on Form 485BPOS contracts, to evaluate LAW's performance.
- The paper emphasizes the extensibility of the LAW framework, allowing new tools and agents to be integrated seamlessly for additional tasks or domains.
- The paper introduces text agents like the Summary Agent and Comparison Agent to analyze and condense information from lengthy legal documents.
- The paper discusses how LAW utilizes modular workflows to address the challenges of long and unstructured legal texts, splitting and processing content into manageable chunks.
- The paper details the use of FlowMind's framework for orchestrating tools and agents, providing a scalable and reusable approach to AI-driven legal workflows.
- The paper outlines potential future applications of LAW, such as adapting it for non-English contracts and expanding its use to other financial and regulatory domains.
Cited By
Quotes
Abstract
Legal contracts in the custody and fund services domain govern critical aspects such as key provider responsibilities, fee schedules, and indemnification rights. However, it is challenging for an off-the-shelf LLM to ingest these contracts due to the lengthy unstructured text streams, limited LLM context windows, and complex legal jargon. To address these challenges, we introduce LAW (Legal Agentic Workflows for Custody and Fund Services Contracts). LAW features a modular design that responds to user queries by orchestrating a suite of domain-specific tools and text agents. Our experiments demonstrate that LAW, by integrating multiple specialized agents and tools, significantly outperforms the baseline. LAW excels particularly in complex tasks such as calculating a contract's termination date, surpassing the baseline by 92.9 percentage points. Furthermore, LAW offers a cost-effective alternative to traditional fine-tuned legal LLMs by leveraging reusable, domain-specific tools.
References
;
| Author | volume | Date Value | title | type | journal | titleUrl | doi | note | year | |
|---|---|---|---|---|---|---|---|---|---|---|
| 2024 LAWLegalAgenticWorkflowsforCust | Manuela Veloso William Watson Nicole Cho Nishan Srishankar Zhen Zeng Lucas Cecchi Daniel Scott Suchetha Siddagangappa Rachneet Kaur Tucker Balch | LAW: Legal Agentic Workflows for Custody and Fund Services Contracts | 2024 |