README.md

metadata

title: IndiaFinBench
emoji: 📜
colorFrom: red
colorTo: yellow
sdk: docker
app_port: 7860
pinned: false
license: mit
short_description: LLM benchmark for Indian financial regulation

IndiaFinBench

The first evaluation benchmark for large language model performance on Indian financial regulatory text.

406 expert-annotated questions over 192 SEBI & RBI regulatory documents (1992–2026) · 12 frontier models evaluated · hybrid FAISS + BM25 retrieval with Recall@5 = 0.785.

This Space hosts the live research site: the full leaderboard with statistical tier analysis, a dataset explorer, a live hybrid-RAG demo over the regulatory corpus, and model submission.

• Dataset: Rajveer-code/IndiaFinBench (CC BY 4.0)
• Code & paper: github.com/Rajveer-code/IndiaFinBench (MIT)
• Author: Rajveer Singh Pall — rajveerpall04@gmail.com