LLM Parameters for Math Across Languages: Shared or Separate?

A new study from arXiv reveals that large language models use partially shared parameters for mathematical reasoning across languages, with the strongest overlap in intermediate layers. English yields the largest set of math-relevant parameters, while lower-resource languages show smaller sets, indicating systematic language-dependent differences.

arXiv:2606.18453v1 Announce Type: new Abstract: Large language models LLMs exhibit substantial cross-lingual variation in mathematical reasoning performance, but it remains unclear whether these differences reflect language-specific parameters or a shared mechanism that manifests differently by language. We present a cross-lingual mechanistic analysis of mathematical reasoning in LLMs, enabling us to localize and compare model parameters that support mathematical reasoning across languages. We find that the extracted math-associated parameters exhibit partial cross-lingual overlap, with the strongest overlap concentrated in intermediate model layers. We further observe that English consistently produces the largest set of math-relevant parameters, whereas lower-resource languages reveal smaller sets of relevant parameters. These results suggest that math-related behavior in multilingual LLMs is neither fully language-invariant nor fully language-specific, but instead exhibits partial cross-lingual parameter overlap with systematic language-dependent differences.