提升多语言事实记忆一致性，新方法在跨语言生成中表现更优
PolyFact: Comparing Consistency-Driven Post-training Methods for Cross-Lingual Factual Recall
- 基于跨语言奖励池的GRPO方法增强事实一致性
- 新方法在未见语言和自由生成中准确率最高，达72.3%
- 适合关注多语言知识推理与模型泛化能力的研究者
大型语言模型虽在英语数据上训练，但跨语言事实表达常不一致。我们构建了包含60,000个基于Wikidata的事实、覆盖12种语言类型的并行多语言问答数据集PolyFact，提出一致性驱动的GRPO方法，结合跨语言奖励池。在OLMo-2-1124-7B和Qwen-2.5-7B上对比SFT、DCO、CM-Align等基线方法。结果表明：SFT提升分布内准确率但不改善一致性；DCO增强一致性但无法迁移至自由生成；我们的GRPO变体在多语言基础模型上的自由生成和未见语言中表现最佳，准确率达72.3%。轻量持续预训练（CPT）对单语模型略有帮助，但损害多语模型性能。机制分析显示GRPO降低语言特异性，促进跨语言表征共享。代码、模型与数据集已公开。
原文摘要 · Abstract (English)
Large language models (LLMs) trained predominantly on English data encode substantial world knowledge, yet often fail to express it reliably in other languages, a phenomenon known as cross-lingual factual inconsistency. To study this, we introduce PolyFact, a fully parallel multilingual factual QA dataset of 60K Wikidata-grounded facts across 12 typologically diverse languages, and propose consistency-driven GRPO with cross-lingual reward pooling. We compare our method against supervised fine-tuning (SFT) and the consistency-enhancement baselines DCO and CM-Align on OLMo-2-1124-7B and Qwen-2.5-7B, and analyze whether light continual pretraining (CPT) on parallel data provides a useful foundation for post-training. No single method dominates: SFT maximises in-distribution accuracy but not consistency, DCO yields the strongest consistency gains but fails to transfer to free-form generation, and our GRPO variant achieves the strongest transfer to free-form recall and unseen languages on the multilingual base model. CPT mildly aids monolingual models but harms multilingual ones. Mechanistic analyses suggest that GRPO is associated with reduced language specialization, consistent with greater sharing of representations across languages. We release our code, models, and dataset publicly.
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。