作者位次 #3/3ACL 2026Findings· Findings of ACL
F rame N et-Cultures: A Benchmark for Evaluating LLM s via Cross-Cultural Frame Semantics
Neda Jamshidi · Anders Søgaard · Monica Bianchini
摘要
Large language models (LLMs) exhibit cultural biases, yet existing benchmarks rely on closed-form, domain-specific questionnaires. We introduce FRAMENET-CULTURES, a benchmark for evaluating cultural alignment in LLMs based on Fillmore-style frame semantics. Using the EveryCulture encyclopedia, we...