Hinata Tezuka


2025

pdf bib
The Transfer Neurons Hypothesis: An Underlying Mechanism for Language Latent Space Transitions in Multilingual LLMs
Hinata Tezuka | Naoya Inoue
Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing

Recent studies have suggested a processing framework for multilingual inputs in decoder-based LLMs: early layers convert inputs into English-centric and language-agnostic representations; middle layers perform reasoning within an English-centric latent space; and final layers generate outputs by transforming these representations back into language-specific latent spaces.However, the internal dynamics of such transformation and the underlying mechanism remain underexplored.Towards a deeper understanding of this framework, we propose and empirically validate **The Transfer Neurons Hypothesis**: certain neurons in the MLP module are responsible for transferring representations between language-specific latent spaces and a shared semantic latent space.Furthermore, we show that one function of language-specific neurons, as identified in recent studies, is to facilitate movement between latent spaces.Finally, we show that transfer neurons are critical for reasoning in multilingual LLMs