发现大模型学两跳推理需记两次事实，小模型易陷入死记硬背陷阱。
Examining Two Hop Reasoning Through Information Content Scaling
- 通过容量缩放实验，揭示两跳推理需记忆双倍事实
- 小模型在特定数据下会死记答案而非组合推理
- 为模型能力评估提供新视角，适合研究推理机制者
先前研究发现，Transformer 在学习隐含两跳问题（如“Bob 的母亲的老板是谁？”）时表现不一。本文通过考察模型规模与两跳问答数据集学习能力的关系，研究其原因。结果表明，容量缩放和泛化均支持：隐含两跳问答需模型重复学习每个事实，而带思维链的两跳问答则无需。此外，我们发现通过调整数据集参数，可使极小模型陷入仅记忆答案的困境，即使组合推理更优。研究显示，容量缩放可补充现有可解释性方法，但应用中仍存在挑战。
原文摘要 · Abstract (English)
Prior work has found that transformers have an inconsistent ability to learn to answer latent two-hop questions -- questions of the form "Who is Bob's mother's boss?" We study why this is the case by examining how transformers' capacity to learn datasets of two-hop questions and answers (two-hop QA) scales with their size, motivated by prior work on transformer knowledge capacity for simple factual memorization. We find that capacity scaling and generalization both support the hypothesis that latent two-hop QA requires transformers to learn each fact twice, while two-hop QA with chain of thought does not. We also show that with appropriate dataset parameters, it is possible to "trap" very small models in a regime where they memorize answers to two-hop questions independently, even though they would perform better if they could learn to answer them with function composition. Our findings show that measurement of capacity scaling can complement existing interpretability methods, though there are challenges in using it for this purpose.
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。