让大模型自动匹配提问语言，提升多语种推理准确率与效率。
AdaMame: A Training Recipe for Adaptive Multilingual Reasoning
- 分两阶段训练：先用五种语言的自然推理数据微调，再用动态对齐策略优化语言适配。
- 在12种语言上测试，比基线模型在准确率、语言一致性和生成效率上全面领先。
- 特别适合低资源语言和跨域场景，避免中英文混杂与多余 token 消耗。
大型推理模型（LRMs）在英语任务中表现优异，但在非英语提问时常出现语言坍缩现象。现有基于强化学习的方法通常添加二元语言一致性奖励，但会牺牲准确性、导致推理过程中的语言切换，并增加令牌使用量。本文提出 AdaMame，一种面向多语种数学推理的两阶段训练方案，通过自适应对齐推理语言与查询语言，在不损失准确率的前提下解决上述问题。第一阶段采用监督微调（SFT），在五种语言的自然推理轨迹上进行微调，建立多语言推理能力；第二阶段引入 AdaMame-GRPO，基于组相对策略优化（GRPO）改进，使查询条件下的对齐因子随训练进程逐步增长，引导模型先探索多种语言推理路径，再聚焦于查询语言的高效表达。在两个基准测试、两种模型和12种语言上的评估显示，AdaMame-GRPO 在推理准确率、语言一致性与令牌效率三方面均达到帕累托最优，尤其在跨域及低资源语言上优势显著。
原文摘要 · Abstract (English)
While Large Reasoning Models (LRMs) show strong performance in English, they often fail to reason in the language of the query, a phenomenon known as language collapse. Existing RL-based fixes typically add a binary language fidelity reward to the accuracy objective, yet still incur trade-off in accuracy, mid-trace code-switching, and excessive token usage. In this work, we propose AdaMame, a two-stage training recipe for multilingual mathematical reasoning that addresses these limitations by adaptively aligning the reasoning language to the query language without compromising accuracy. The first SFT stage fine-tunes on naturally occurring reasoning traces across five languages to establish multilingual reasoning capability. In the subsequent RL stage, we introduce AdaMame-GRPO, an adaptation of Group Relative Policy Optimization (GRPO) in which a query-conditioned alignment factor grows progressively during training, guiding the model to first explore diverse reasoning languages before exploiting reasoning in the query language. Evaluated across two benchmarks, two LRMs, and 12 languages, AdaMame-GRPO achieves Pareto-optimal performance across reasoning accuracy, language fidelity, and token efficiency over all baselines, with the strongest gains on out-of-domain, lower-resource languages.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。