登录
洞察 - Language Research
LIEDER: Linguistically-Informed Evaluation for Discourse Entity Recognition at Yale University
Language models exhibit knowledge of EXISTENCE, UNIQUENESS, and PLURALITY but lack understanding of NOVELTY in DE recognition.
MELA: Multilingual Evaluation of Linguistic Acceptability
MELA introduces a multilingual benchmark for linguistic acceptability, highlighting the importance of in-language training data for cross-lingual transfer and syntax-related tasks.
Correlation Between Language Distance and Cross-Lingual Transfer in Multilingual Space
Linguistic features impact cross-lingual transfer performance and representation spaces in multilingual models.
MELA: Multilingual Evaluation of Linguistic Acceptability
MELA introduces a multilingual benchmark for linguistic acceptability judgment, showcasing the importance of in-language training data and its impact on syntax-related tasks.
