BFSI-Bench
Dataset ground-truth.in

190 English questions · CC BY 4.0

BFSI-Bench

BFSI-Bench tests how well language models answer questions about India’s banking, financial services, and insurance (BFSI) rules.

Hugging Face dataset

Introduction

BFSI-Bench is a 190-item English test set for question answering on India’s banking, financial services, and insurance rules. Domain experts wrote the questions and reference answers and checked them against official sources such as the Reserve Bank of India (RBI), the Securities and Exchange Board of India (SEBI), and the Insurance Regulatory and Development Authority of India (IRDAI).

Scoring compares a model’s reply to short, checkable facts drawn from those sources, not only to the prose gold answer. The public download includes source links and quoted spans. It does not include the full source documents. You can inspect every item on the samples page.

Categories

Each item belongs to one of five categories. The count is the number of test items in that category.

Jurisdiction · 71 questions Jurisdiction-Aware Compliance

The model should apply Indian rules, or ask which country the user means, rather than assuming U.S. or EU law.

Numerical · 36 questions Numerical Reasoning

The model should compute Indian finance quantities correctly, including EMI, interest, and tax deducted at source (TDS).

Temporal · 18 questions Temporal Logic

When the question does not name a year, the model should use the circular or limit that applies now.

Red-team · 29 questions Red-Team Evasion

The model should refuse illegal financial workarounds and point to legitimate channels instead.

Groundedness · 36 questions Groundedness

The answer should stay faithful to the circular, scheme, or policy named in the question.

Item fields

Each row in the test split has these fields.

Field Description
Identifier Stable public id, such as bfsi-jur-001
Question The prompt we send the model
Expert answer The reference answer written by a domain expert
Facts Short claims we check in a model’s reply. Some are marked required.
Sources Official page links, a short note, quotes from those pages, and a publication date when we have one

Citation

@dataset{groundtruth2026bfsibench,
  title={BFSI-Bench: Fact-Seeking Eval for Indian BFSI},
  author={GroundTruth},
  year={2026},
  url={https://huggingface.co/datasets/ground-truth/bfsi-bench},
  license={CC-BY-4.0}
}

Questions about this work: miroojin@ground-truth.in