arXivDaily arXiv每日学术速递 周一至周五更新
北美退伍军人节(Veterans Day),12 日无更新
arXiv 2606.02465cs.CLcs.AI
收藏

学习何时翻译以实现多语言推理

Learning When to Translate for Multilingual Reasoning

Graduate School of Artificial Intelligence(人工智能研究生院);Department of Computer Science & Engineering(计算机科学与工程系);POSTECH

Deokhyung Kang, Hyounghun Kim

AI总结:

提出Luar框架,通过强化学习训练推理语言模型在直接理解不可靠时选择性调用翻译,从而缩小多语言推理差距。

摘要:

推理语言模型(RLMs)在复杂推理任务上表现出色,但仍存在显著的多语言推理差距,这主要源于非英语输入中的语言理解失败。英语翻译可以通过将非英语输入转换为RLMs更可靠解释的形式来缓解这些失败,但当模型能够从原始查询中可靠推理时,翻译每个输入是不必要的。为应对这一挑战,我们提出Luar,一种语言理解边界感知的强化学习框架,训练RLMs在直接理解不可靠时选择性调用翻译。Luar训练模型在直接解决原始输入和对其英语翻译进行推理之间做出选择,仅在翻译增强推理预期显著优于直接推理时鼓励翻译。在多语言推理基准测试中,Luar优于标准GRPO和其他基于训练的基线,在低资源语言上尤其获得巨大提升。进一步分析表明,Luar在直接推理足够的情况下避免不必要的翻译,同时将其翻译调用行为扩展到未见过的低资源语言。总之,我们的工作提出了一种选择性多语言推理方法:RLMs可以学习仅在直接理解不可靠时调用翻译。该项目将在https://github.com/deokhk/LUAR公开。

英文摘要:

Reasoning language models (RLMs) achieve strong performance on complex reasoning tasks, but still exhibit substantial multilingual reasoning gaps, largely due to language-understanding failures when interpreting non-English inputs. Translation into English can mitigate these failures by expressing non-English inputs in a form that RLMs can more reliably interpret, yet translating every input incurs unnecessary inference cost when the model can reason reliably from the original query. To address this challenge, we propose Luar, a Language Understanding Boundary-aware Reinforcement Learning framework that trains RLMs to selectively invoke translation when their direct understanding is unreliable. Luar trains the model to either solve the original input directly or reason over its English translation, encouraging translation only when translator-augmented reasoning is expected to substantially outperform direct reasoning. Across multilingual reasoning benchmarks, Luar outperforms training-based baselines, with particularly large gains on low-resource languages. On Qwen3-4B, it improves average accuracy by 8.4 percentage points over standard GRPO. Further analysis shows that Luar concentrates translator calls on queries where translation benefits reasoning, helping to reduce inference cost. The project will be made publicly available at https://github.com/deokhk/LUAR.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑