LexChronos: An Agentic Framework for Structured Event Timeline Extraction in Indian Jurisprudence

Anka Chandrahas Tummepalli; Preethu Rose Anish

cs.CLMar 2, 2026

LexChronos: An Agentic Framework for Structured Event Timeline Extraction in Indian Jurisprudence

Authors: Anka Chandrahas Tummepalli, Preethu Rose Anish

Organizations: TCS Research

Abstract

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.

Figures & tables

Judgment Component	LES Coverage	Justification of Coverage
Facts	Timestamp, Event	Factual elements are encoded as Events with corresponding Timestamps establishing the case chronology.
Issues	Event	The framing of issues is a critical procedural Event which dictates the focus of the subsequent legal analysis.
Petitioner’s Arguments	Event, Precedent	Arguments constitute procedural Events ; their basis relies heavily on Precedent or statutory references.
Respondent’s Arguments	Event, Precedent	Similar to petitioner’s claims, these are timed procedural Events grounded in Precedent .
Analysis of the Law	Event, Precedent	The application of law is a judicial Event and is intrinsically linked to the Precedent being discussed.
Precedent Analysis	Event, Precedent, Judge	This analytical Event explicitly involves the cited Precedent and the authoritative Judge performing the analysis.

Table 1: Mapping core Judgment components to LES

Model	Avg Length of Judgments	Avg Events per Case	Avg Precedents per Case	Unique Vocab Size
DeepSeek-R1	1010.96	27	6	34623
GPT-4	552.92	19	3	17080

Table 2: Structural and linguistic comparison of DeepSeek R1 and GPT-4 judgments

Core Judgment Component with Judgment text
Facts
“The genesis of the dispute traces back to January 2020, when Proxima Inc. acquired a substantial stake in Zenith Corp, leading to disputes concerning corporate governance and the rights of minority shareholders.”
Issues
“..the principal issues discussed revolved around corporate governance, the rights of minority shareholders, and the extent of judicial oversight permissible over corporate decisions.”
Petitioner’s Arguments
“In February 2020, the minority shareholders of Zenith Corp alleged oppression and mismanagement by the board members of Proxima Inc., leading to the filing of a complaint…”

Table 3: Structural Component Mapping for Synthetic Judgment Text

Figure 1: Reverse-Engineering Pipeline for Gold-Standard Event timeline Dataset

Figure 2: The LexChronos Framework for Iterative Extraction and Feedback-Driven Refinement

LLM	Bert Precision	Bert Recall	Bert F1
Llama 3.2 3B Instruct	0.8450	0.8336	0.8390
Gemma 3 1B IT	0.8237	0.7992	0.8113
Gemma 2 2B IT	0.7809	0.7957	0.7880
Llama 3.2 1B Instruct	0.6058	0.6054	0.6052
DeepSeek R1 Distill Qwen 1.5B	0.5548	0.5135	0.5276
Qwen 2.5 3B	0.5628	0.4467	0.4752

Table 4: Extraction Agent performance on

n_{test}

Feedback Configuration	Bert Precision	Bert Recall	Bert F1
Self Feedback	0.8681	0.8371	0.8523
Gemma 2 2B IT	0.8918	0.8590	0.8751
Gemma 3 1B IT	0.8664	0.8348	0.8502
Llama 3.2 3B Instruct	0.8665	0.8351	0.8505
Llama 3.2 1B Instruct	0.8661	0.8344	0.8499

Table 5: Feedback Agent Impact on Extraction Quality

LLM	$S_{\textbf{{unstructured}}}$	$S_{\textbf{{structured}}}$
Llama 3.1 8B Instruct	55	145
Gemma 2 9B IT	50	150

Table 6: GPT-4 Preference for Structured Input Summaries (

n_{sum}

)

Appendix figures & tables1 asset

Supplementary material from the paper’s appendix.

Appendix

Case Categories
Constitutional Law	Civil Law
Banking and Finance Law	Education Law
Intellectual Property Rights (IPR)	Taxation Law
Labor and Employment Law	Family Law
Consumer Protection Law	Cyber Law
Real Estate and Property Law	Criminal Law

Appendix

Table 7: List of Case Categories

Legal NLP

Legal natural language processing (NLP) develops methods for analyzing and generating statutes, cases, contracts, and legal advice.

See all 102 papers

Vichara: Appellate Judgment Prediction and Explanation for the Indian Judicial System
Feb 20, 2026Pavithra PM Nair, Preethu Rose AnishLegal Judgment Prediction Legal NLP
TCS Research
NyayaMind- A Framework for Transparent Legal Reasoning and Judgment Prediction in the Indian Legal System
Apr 10, 2026Parjanya Aditya Shukla, Shubham Kumar Nigam, Debtanu Datta +5Legal Judgment Prediction Legal Reasoning
IIT Kanpur, India · University of Birmingham, Dubai, United Arab Emirates · IIT Kharagpur, India +2
NyayaAI: An AI-Powered Legal Assistant Using Multi-Agent Architecture and Retrieval-Augmented Generation
May 11, 2026Deepanshu, Divi Saxena, Deepali Rana +2Multi-Agent LLM Systems Legal IR
School of Computer Science UPES, Dehradun, India

Feb 20, 2026J/K move · Enter open · S save

Vichara: Appellate Judgment Prediction and Explanation for the Indian Judicial System

Pavithra PM Nair, Preethu Rose Anish

TCS Research

Legal Judgment Prediction Legal NLP

Open paper AskPDF arXiv

In jurisdictions like India, where courts face an extensive backlog of cases, artificial intelligence offers transformative potential for legal judgment prediction. A critical subset of this backlog comprises appellate cases, which are formal decisions issued by higher courts reviewing the rulings of lower courts. To this end, we present Vichara, a novel framework tailored to the Indian judicial system that predicts and explains appellate judgments. Vichara processes English-language appellate case proceeding documents and decomposes them into decision points. Decision points are discrete legal determinations that encapsulate the legal issue, deciding authority, outcome, reasoning, and temporal context. The structured representation isolates the core determinations and their context, enabling accurate predictions and interpretable explanations. Vichara's explanations follow a structured format inspired by the IRAC (Issue-Rule-Application-Conclusion) framework and adapted for Indian legal reasoning. This enhances interpretability, allowing legal professionals to assess the soundness of predictions efficiently. We evaluate Vichara on two datasets, PredEx and the expert-annotated subset of the Indian Legal Documents Corpus (ILDC_expert), using four large language models: GPT-4o mini, Llama-3.1-8B, Mistral-7B, and Qwen2.5-7B. Vichara surpasses existing judgment prediction benchmarks on both datasets, with GPT-4o mini achieving the highest performance (F1: 81.5 on PredEx, 80.3 on ILDC_expert), followed by Llama-3.1-8B. Human evaluation of the generated explanations across Clarity, Linking, and Usefulness metrics highlights GPT-4o mini's superior interpretability.

Abstract

Legal NLP

Explore similar work

Vichara: Appellate Judgment Prediction and Explanation for the Indian Judicial System

NyayaMind- A Framework for Transparent Legal Reasoning and Judgment Prediction in the Indian Legal System

NyayaAI: An AI-Powered Legal Assistant Using Multi-Agent Architecture and Retrieval-Augmented Generation