Benchmark RadarL3Cube-IndicQuest v2
L3Cube-IndicQuest v2 is a multilingual QA benchmark for evaluating India-specific factual knowledge of LLMs. It contains 3,471 curriculum-grounded English QA pairs translated into 19 Indic languages, totaling 69,420 pairs across 20 languages.
- Released
- 2026-08-16
- Readiness
- Paper only
- Primary field
- General AI
Why it matters
There is a need for benchmarks evaluating factual knowledge in Indic languages. This benchmark provides a large-scale gold-standard dataset with multiple evaluation protocols, showing consistent rankings across judges.
Motivation
We present L3Cube-IndicQuest v2, a large-scale gold-standard multilingual question-answering benchmark for evaluating the India-specific factual knowledge of Large Language Models (LLMs).
Primary resources
Benchmark Radar records only publicly supported details and links back to primary sources for verification.