Suchir Salhan

Monolingual models can achieve cross-lingual alignment without joint training, revealing the power of linguistic structure over shared parameters.

Language-adaptive tokenization can significantly enhance morphological alignment without the need for separate vocabularies for each language.

Statistical regularities in phoneme frequency distributions, previously thought to arise from optimization, may instead be natural consequences of diachronic sound change.

Phoneme frequency distributions across languages aren't random noise; they follow predictable patterns governed by inventory size and linguistic constraints, offering a unified information-theoretic account.

BabyLM 2026 seeks to push the boundaries of data-efficient and cognitively plausible language models, now with a multilingual twist.