ACL 2025: Search-based multilingual LLM evaluation that discovers English-correct, target-language-wrong failures; code and 6,713-pair dataset across 16 languages.
multilingual-evaluation low-resource-languages llm-evaluation cross-lingual-evaluation multilingual-llm adversarial-evaluation acl-2025 failure-discovery cross-lingual-consistency cross-lingual-weakness automatic-probing multilingual-post-training
-
Updated
Sep 21, 2026 - Python