adalat-ai/indian-legal-exam-benchmark

Indian Law Entrance Exam Papers Dataset

A comprehensive, research-ready collection of multiple-choice questions from key Indian law entrance exams, including CLAT (UG & PG) and Delhi Judicial Service papers. This dataset is ideal for legal education, NLP benchmarking, and educational technology.

Paper: Are LLMs Court-Ready? Evaluating Frontier Models on Indian Legal Reasoning

---

Overview

This dataset provides 6,218 exam questions from 38 law entrance papers conducted between 2008 and 2025. The dataset is organized into three subsets:

• CLAT_UG: CLAT Undergraduate (LLB)
• CLAT_PG: CLAT Postgraduate (LLM)
• DJSS_DHJS: Delhi Judicial Service and Higher Judicial Service Exams

---

Exam Descriptions & Subject Coverage

Exam Type	Description	Subjects Covered
CLAT UG	Entrance exam for LLB programs at Law Universities. Assesses comprehension, reasoning, legal aptitude, and basic mathematics.	English Language; Current Affairs / GK; Legal Reasoning; Logical Reasoning; Quantitative Techniques.
CLAT PG	National entrance for LLM admissions. Focuses on core areas of law, analytical ability, legal principles, and advanced reading.	Core LL.B. subjects: Constitutional Law; Jurisprudence; Administrative Law; Contract; Torts; Family; Criminal; Property; Company; Public International Law; Tax; Environmental; Labour / Industrial Law.
DJS/DHJS	Competitive exams for Delhi Civil Judge and Higher Judicial Service posts. Emphasizes statutory law, procedural rules, and legal reasoning.	Civil Law, Criminal Law, Evidence, CPC, CrPC, Constitutional Law, Contract, Local Delhi Laws, Legal Aptitude

---

Dataset Structure

• Questions: 6,218
• Papers: 38
• Years: 2008–2025
• Fields per entry:
  • question_text: The question
  • options: List of 4 options [A, B, C, D]
  • answer: Correct answer label (A, B, C, or D)
  • source_paper: Exam identifier (e.g., "clat_2024", "djs_2019_prelims")

---

Loading the Dataset

from datasets import load_dataset

clat_ug = load_dataset("adalat-ai/indian-legal-exam-benchmark", "clat_ug")
clat_pg = load_dataset("adalat-ai/indian-legal-exam-benchmark", "clat_pg")
djs_dhjs = load_dataset("adalat-ai/indian-legal-exam-benchmark", "djs_dhjs")

for item in clat_ug["train"]:
    print(f"Question: {item['question_text']}")
    print(f"Options: {item['options']}")
    print(f"Answer: {item['answer']}")
    print(f"Source: {item['source_paper']}")
    break

---

Applications

• Legal Education: Exam preparation and practice for law students.
• NLP/AI Research: Benchmark for legal question answering, context comprehension, and language model assessment in law.
• EdTech: Building adaptive legal learning and test simulation platforms.

---

Dataset Statistics

Exam Type	Papers	Questions	Years
CLAT UG	18	3,154	2008–2025
CLAT PG	7	814	2019–2025
DJS/DHJS	13	2,250	2011–2023
Total	38	6,218	2008–2025

---

---