细粒度损失截断的优势研究：以摘要生成事实性为例
On the Benefits of Fine-Grained Loss Truncation: A Case Study on Factuality in Summarization
AI总结：
针对文本摘要模型的幻觉问题，研究发现现有Loss Truncation方法因假设局限效果有限，提出基于实体词级NLL的细粒度损失与数据清洗策略，有效减少部分数据集上的幻觉。
摘要：
文本摘要与简化是应用最广泛的AI应用之一，但这类任务的模型常出现幻觉问题，该问题可能源于未对齐数据的训练。解决此问题的一种高效方法是Loss Truncation（LT，损失截断）（Kang和Hashimoto，2020），该方法修改标准对数损失，在训练过程中自适应移除噪声样本。但我们发现，仅使用LT会在多个数据集上产生大量幻觉实体。为理解并优化LT的性能，我们研究了事实性样本与非事实性样本之间底层损失的表现。我们证明，当“噪声目标具有更高负对数似然（NLL）损失”这一基本假设不成立时，LT的性能会受限，同时发现实体层面的词级NLL能为区分事实性提供更好的信号。我们据此提出细粒度NLL损失与细粒度数据清洗策略，在部分数据集上观察到幻觉减少的效果提升。我们的工作可在https://github.com/yale-nlp/fine-grained-lt获取。
英文摘要：
Text summarization and simplification are among the most widely used applications of AI. However, models developed for such tasks are often prone to hallucination, which can result from training on unaligned data. One efficient approach to address this issue is Loss Truncation (LT) (Kang and Hashimoto, 2020), an approach to modify the standard log loss to adaptively remove noisy examples during training. However, we find that LT alone yields a considerable number of hallucinated entities on various datasets. We study the behavior of the underlying losses between factual and non-factual examples, to understand and refine the performance of LT. We demonstrate that LT's performance is limited when the underlying assumption that noisy targets have higher NLL loss is not satisfied, and find that word-level NLL among entities provides better signal for distinguishing factuality. We then leverage this to propose a fine-grained NLL loss and fine-grained data cleaning strategies, and observe improvements in hallucination reduction across some datasets. Our work is available at https://https://github.com/yale-nlp/fine-grained-lt.
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。
