PCS transfers English reasoning to other languages in LRMs via code-switched SFT initialization followed by step-level RL curriculum that progressively increases target-language ratio, narrowing the performance gap without stronger models.
arXiv preprint arXiv:2511.23231 , year=
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Efficient Multilingual Reasoning Transfer via Progressive Code-Switching
PCS transfers English reasoning to other languages in LRMs via code-switched SFT initialization followed by step-level RL curriculum that progressively increases target-language ratio, narrowing the performance gap without stronger models.