ACL2026
Why Do Multilingual Reasoning Gaps Emerge in Reasoning Language Models?
Deokhyung Kang, Seonjeong Hwang, Daehui Kim, Hyounghun Kim, Gary Lee
被引用 3 次
摘要
Reasoning language models (RLMs) achieve strong performance on complex reasoning tasks, yet they still exhibit a multilingual reasoning gap, performing better in high-resource languages than in low-resource ones. While recent efforts have been made to address this gap, its underlying causes remain largely unexplored. In this work, we show that this gap primarily stems from failures in language understanding-specifically, the model's inability to translate multilingual inputs into the language dominating its reasoning traces (typically English). As identifying understanding failures can enable targeted mitigation of the gap, we evaluate a range of detection methods and find that understanding failures are detectable to a meaningful extent, with supervised approaches performing best. Building on this, we propose Selective Translation, a strategy that incorporates an English translation into the initial reasoning trace when an understanding failure is detected. Experimental results using Qwen3-4B show that Selective Translation substantially bridges the multilingual reasoning gap, achieving near full-translation performance while translating only about 20% of inputs. Together, our results show that failures in language understanding are the primary driver of the multilingual reasoning gap and can be detected and selectively mitigated, clarifying its origin and suggesting a path toward more equitable multilingual reasoning. 1 1 Our code and data are publicly available at here. Reasoning Trace: The question is in Swahili, so I try to translate it. "Kikapu kina machungwa 25 na 1 miongoni mwao limeharibika... Maybe it's about a group of people or animals?…20% are injured, 2 are sick, and the rest are healthy…This is confusing. Maybe the problem is mistranslated or I misunderstand it.… Eventually, if total is 25, 20% is 5, plus 2 sick = 7, so 25 -7 = 18 healthy." ? Question: Kikapu kina machungwa 25 na 1 miongoni mwao limeharibika, 20% hayajaiva, 2 ni chachu na mengine ni mazuri. Machungwa mangapi ni mazuri? (A basket contains 25 oranges among which 1 is bad, 20% are unripe, 2 are sour and the rest are good. How many oranges are good?)