探测大模型内部表征，发现其推理能力与表面表现不一致。
When Models Know More Than They Say: Probing Analogical Reasoning in LLMs
- 通过探测模型内部表示，对比其真实理解与提示响应
- 开源模型在隐喻类类比上探测准确率显著高于提示表现
- 揭示提示工程可能无法充分激活模型潜在认知能力
类比推理是叙事理解的核心认知能力。尽管大模型在表面和结构线索一致时表现良好，但在需要深层抽象的隐性类比任务中表现不佳，表明其在抽象与泛化方面存在局限。本文对比了模型探测表征与提示行为在识别叙事类比上的表现，发现：对于修辞性类比，探测性能显著优于提示；而对于叙事类比，两者均表现低下且相近。这表明模型内部表示与提示行为之间的关系具有任务依赖性，提示可能未能有效访问模型内部已有的知识。
原文摘要 · Abstract (English)
Analogical reasoning is a core cognitive faculty essential for narrative understanding. While LLMs perform well when surface and structural cues align, they struggle in cases where an analogy is not apparent on the surface but requires latent information, suggesting limitations in abstraction and generalisation. In this paper we compare a model's probed representations with its prompted performance at detecting narrative analogies, revealing an asymmetry: for rhetorical analogies, probing significantly outperforms prompting in open-source models, while for narrative analogies, they achieve a similar (low) performance. This suggests that the relationship between internal representations and prompted behavior is task-dependent and may reflect limitations in how prompting accesses available information.
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。