Indian Legal Data V2: 171,640 Instruction-Response Pairs for Legal Language Models

171,640 high-quality instruction–response pairs derived from Indian legal texts, focusing on statutory interpretation and structured legal explanations. The dataset, created by kaushik-harsh-99, is designed for instruction tuning of language models and represents a significant scale upgrade from version 1, which contained 33,077 samples. The dataset page was last updated on 2026-05-01.

Use Cases

• Instruction tuning of language models based on legal reasoning patterns mentioned in the description
• Benchmarking model performance on structured legal explanation tasks based on the described instruction–response format
• Training models for statutory interpretation based on the focus on Indian legal texts
• Improving model clarity and structure in text generation based on the dataset's design emphasis

Strengths

• Contains 171,640 samples, a significant increase from the 33,077 samples in version 1
• Designed with an emphasis on clarity, structure, and legal reasoning patterns for instruction tuning
• Derived from Indian legal texts, providing domain-specific content

Limitations

• Column-level documentation is absent; field semantics must be inferred after download
• Row count is unknown, which may limit suitability assessment
• Description metadata is limited; actual data quality requires manual inspection after download

Provenance

Source

Derived from Indian legal texts, as stated in the description.

Collection Method

Likely involves processing and structuring legal texts into instruction–response pairs, though the exact method is not detailed.

Freshness

Last updated 2026-05-01 17:52:40; freshness should be verified

Geography

India, based on the description's focus on Indian legal texts.