Search papers, labs, and topics across Lattice.
The paper introduces LexChronos, an agentic framework for extracting structured event timelines from Indian Supreme Court judgments to improve legal reasoning tasks. It uses a dual-agent architecture with a LoRA-tuned extraction agent and a pre-trained feedback agent to iteratively identify and refine events. To overcome data scarcity, the authors create a synthetic dataset of 2000 samples with gold-standard event annotations, achieving a BERT-based F1 score of 0.8751, and demonstrate that GPT-4 prefers these structured timelines for legal text summarization.
LLMs prefer structured timelines extracted by a novel agentic framework over unstructured text for legal summarization, suggesting a path to better legal reasoning.
Understanding and predicting judicial outcomes demands nuanced analysis of legal documents. Traditional approaches treat judgments and proceedings as unstructured text, limiting the effectiveness of large language models (LLMs) in tasks such as summarization, argument generation, and judgment prediction. We propose LexChronos, an agentic framework that iteratively extracts structured event timelines from Supreme Court of India judgments. LexChronos employs a dual-agent architecture: a LoRA-instruct-tuned extraction agent identifies candidate events, while a pre-trained feedback agent scores and refines them through a confidence-driven loop. To address the scarcity of Indian legal event datasets, we construct a synthetic corpus of 2000 samples using reverse-engineering techniques with DeepSeek-R1 and GPT-4, generating gold-standard event annotations. Our pipeline achieves a BERT-based F1 score of 0.8751 against this synthetic ground truth. In downstream evaluations on legal text summarization, GPT-4 preferred structured timelines over unstructured baselines in 75% of cases, demonstrating improved comprehension and reasoning in Indian jurisprudence. This work lays a foundation for future legal AI applications in the Indian context, such as precedent mapping, argument synthesis, and predictive judgment modelling, by harnessing structured representations of legal events.
