KINA: Graduate-Level Multidisciplinary Knowledge Benchmark

KINA (Knowledge Index of Noah's Ark) is a benchmark dataset containing 899 ten-option pseudo-multiple-choice questions. It was created by the 2077AIDataFoundation to evaluate large language models across 12 top-level disciplines, 70 fields, and 261 fine-grained subfields. The dataset was last updated on Hugging Face in May 2026.

Use Cases

• Benchmarking LLM knowledge breadth based on coverage of 261 fine-grained subfields
• Evaluating source-grounded reasoning based on graduate-level questions
• Assessing model performance on high-density, multi-option questions

Strengths

• Contains 899 questions providing a substantial test set
• Covers a broad map of 12 top-level disciplines and 70 fields
• Focuses on graduate-level, source-grounded questions for high-quality evaluation

Limitations

• Column-level documentation is absent; field semantics must be inferred after download
• Row count is unknown, which may limit suitability assessment

Provenance

Source

2077AIDataFoundation

Freshness

Last updated 2026-05-28 09:00:29; freshness should be verified