“迷失于后段”：量化大型语言模型中上下文锚定的框架
"Lost-in-the-Later": Framework for Quantifying Contextual Grounding in Large Language Models
AI总结：
本文提出CoPE框架，发现LLM存在“迷失于后段”的位置偏差，即忽视上下文后部信息，并证明基于上下文知识的提示可改善事实锚定、减少幻觉。
摘要：
大型语言模型能够利用上下文知识和参数化知识，但如何优先考虑并整合这些来源仍未被充分探索。我们引入了CoPE，一个新颖的评估框架，系统性地跨模型和语言衡量上下文知识（CK）和参数化知识（PK）。利用我们的MultiWikiAtomic数据集（包含英语、西班牙语和丹麦语），我们分析了大型语言模型（LLMs）如何在开放式问答中整合上下文、优先处理信息并融入PK。我们的分析揭示了一种我们称之为“迷失于后段”的现象，即LLMs倾向于忽视或降低对给定上下文中较后出现信息的优先级，显示出影响上下文锚定的强烈位置偏差。我们进一步发现，推理模型以及使用思维链（CoT）提示的非推理模型，比未使用CoT的非推理模型更少地使用上下文，并且未能缓解“迷失于后段”效应。特别是，CoT提示导致召回率降低和回答缩短，导致上下文锚定质量下降。基于这些见解，我们设计了基于提示的方法来有效利用输入上下文。将CoPE应用于摘要生成的案例研究表明，基于CK的提示改善了事实锚定并减少了幻觉。
英文摘要：
Large language models are capable of leveraging both contextual and parametric knowledge but how they prioritize and integrate these sources remains underexplored. We introduce CoPE, a novel evaluation framework that systematically measures contextual knowledge (CK) and parametric knowledge (PK) across models and languages. Using our MultiWikiAtomic dataset in English, Spanish, and Danish, we analyze how large language models (LLMs) integrate context, prioritize information, and incorporate PK in open-ended question answering. Our analysis uncovers a phenomenon we call lost-in-the-later, where LLMs tend to overlook or deprioritize information that appears later in a given context, revealing a strong positional bias that affects contextual grounding. We further find that reasoning models, as well as non-reasoning models prompted with chain-of-thought (CoT), use context even less than non-reasoning models without CoT and fail to mitigate the lost-in-the-later effect. CoT prompting, in particular, results in lower recall and shorter responses, leading to degraded contextual grounding. Based on these insights, we design prompt-based methods to effectively leverage input context. A case study applying CoPE to summarization demonstrates that CK-informed prompting improves factual grounding and reduces hallucination.
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。
