04:00
2026-08-18
arxiv.org
large-language-models
Interpretable Cross-Lingual Alignment in Small Language Models: Probing Cultural and Pragmatic Reasoning in Japanese-English Bilingual LLMs
A new study introduces J-PragEval-v0, a minimal-pair benchmark isolating four pragmatic phenomena in Japanese, and finds that in TinySwallow-1.5B, honorific register is linearly decodable in the resid…