Benchmark RadarAI BENCHMARK PROFILE
mmPISA-bench
mmPISA-bench consists of 25 multiple-choice questions from PISA in 43 languages with official and machine translations, used to evaluate LLMs' reasoning across languages.
- Released
- 2026-06-05
- Readiness
- Paper only
- Primary field
- General AI
Why it matters
It addresses multilingual reasoning evaluation but the small scale and focus on proprietary models limit its utility as a general benchmark.
Motivation
We introduce mmPISA-bench, a compact high-quality multilingual reasoning benchmark derived from the OECD Programme for International Student Assessment (PISA).
Primary resources
Benchmark Radar records only publicly supported details and links back to primary sources for verification.