Tag
This paper investigates how multilingual tokenizers handle cross-lingual homographs (identical surface forms with different meanings across languages) and proposes a lightweight language-cue intervention that introduces language-specific characters to reduce token sharing. Experiments show modest improvements in machine translation, particularly with BPE tokenization.
BLOOMZ and mT0 are multilingual models finetuned on the crosslingual task mixture xP3, enabling zero-shot instruction following in dozens of languages.
This paper presents a large-scale evaluation of nine uncertainty estimation methods for LLMs across 22 languages, finding that prompting models to reason in English improves uncertainty estimation for low-resource languages and that the choice of method depends on model scale.
The paper proposes Crosslingual On-Policy Self-Distillation (COPSD), a method to transfer high-resource language reasoning capabilities to low-resource languages using a shared student-teacher architecture. Experiments across 17 African languages show significant improvements in mathematical reasoning and answer-format adherence, outperforming Group Relative Policy Optimization (GRPO).