SYSTEM Cited by 1 source

ECLeKTic benchmark

ECLeKTic (Evaluating Cross-Lingual Knowledge Transfer in LLMs) is a Google Research benchmark for measuring how well an LLM transfers knowledge it has learned in one language to question-answering in another language. The benchmark is introduced on the Google Research blog and cited in the 2026-05-28 I/O 2026 roundup post as part of Google's multilinguality research arc (Source: sources/2026-05-28-google-a-new-era-of-innovation-google-research-at-io-2026).

This is a minimum-viable wiki page anchored to the I/O 2026 post's pointer to the ECLeKTic blog. Architectural / dataset / methodology detail is not in this raw and lives in the linked ECLeKTic blog post and the underlying paper.

Role

ECLeKTic is one of two named benchmarks in the I/O 2026 post's multilinguality discussion (alongside a geographic-localization benchmark at arXiv:2604.19292). Together they support Google's claim that Gemini is "the most widely available AI assistant in the world" — deployed in "more than 70 languages across more than 230 countries."

Seen in

• sources/2026-05-28-google-a-new-era-of-innovation-google-research-at-io-2026 — cited as a multilinguality benchmark for measuring cross-lingual knowledge transfer in LLMs.

Related

• concepts/multilingual-llm-evaluation — the broader evaluation discipline.
• concepts/cross-lingual-knowledge-transfer — the specific phenomenon ECLeKTic measures.
• systems/gemini — the production LLM family this evaluation supports.
• companies/google — operator.