Delving into the Reversal Curse: How Far Can Large Language Models Generalize?

• URL: http://arxiv.org/abs/2410.18808v1
• Date: Thu, 24 Oct 2024 14:55:09 GMT
• Title: Delving into the Reversal Curse: How Far Can Large Language Models Generalize?
• Authors: Zhengkai Lin, Zhihang Fu, Kai Liu, Liang Xie, Binbin Lin, Wenxiao Wang, Deng Cai, Yue Wu, Jieping Ye,
• Abstract summary: A prime example is the recently debated "reversal curse", which surfaces when models, having been trained on the fact "A is B", struggle to generalize this knowledge to infer that "B is A" In this paper, we examine the manifestation of the reversal curse across various tasks and delve into both the generalization abilities and the problem-solving mechanisms of LLMs.
• Score: 40.64539467276017
• License:
• Abstract: While large language models (LLMs) showcase unprecedented capabilities, they also exhibit certain inherent limitations when facing seemingly trivial tasks. A prime example is the recently debated "reversal curse", which surfaces when models, having been trained on the fact "A is B", struggle to generalize this knowledge to infer that "B is A". In this paper, we examine the manifestation of the reversal curse across various tasks and delve into both the generalization abilities and the problem-solving mechanisms of LLMs. This investigation leads to a series of significant insights: (1) LLMs are able to generalize to "B is A" when both A and B are presented in the context as in the case of a multiple-choice question. (2) This generalization ability is highly correlated to the structure of the fact "A is B" in the training documents. For example, this generalization only applies to biographies structured in "[Name] is [Description]" but not to "[Description] is [Name]". (3) We propose and verify the hypothesis that LLMs possess an inherent bias in fact recalling during knowledge application, which explains and underscores the importance of the document structure to successful learning. (4) The negative impact of this bias on the downstream performance of LLMs can hardly be mitigated through training alone. Based on these intriguing findings, our work not only presents a novel perspective for interpreting LLMs' generalization abilities from their intrinsic working mechanism but also provides new insights for the development of more effective learning methods for LLMs.

Related papers

• What Do Language Models Learn in Context? The Structured Task Hypothesis [89.65045443150889]
  Large language models (LLMs) learn a novel task from in-context examples presented in a demonstration, termed in-context learning (ICL) One popular hypothesis explains ICL by task selection. Another popular hypothesis is that ICL is a form of meta-learning, i.e., the models learn a learning algorithm at pre-training time and apply it to the demonstration.
  arXiv  Detail & Related papers  (2024-06-06T16:15:34Z)
• LLMs' Reading Comprehension Is Affected by Parametric Knowledge and Struggles with Hypothetical Statements [59.71218039095155]
  Task of reading comprehension (RC) provides a primary means to assess language models' natural language understanding (NLU) capabilities. If the context aligns with the models' internal knowledge, it is hard to discern whether the models' answers stem from context comprehension or from internal information. To address this issue, we suggest to use RC on imaginary data, based on fictitious facts and entities.
  arXiv  Detail & Related papers  (2024-04-09T13:08:56Z)
• Meaningful Learning: Advancing Abstract Reasoning in Large Language Models via Generic Fact Guidance [38.49506722997423]
  Large language models (LLMs) have developed impressive performance and strong explainability across various reasoning scenarios. Despite this, when tasked with simple questions supported by a generic fact, LLMs often fail to provide consistent and precise answers. This has sparked a vigorous debate about whether LLMs are genuinely reasoning or merely memorizing.
  arXiv  Detail & Related papers  (2024-03-14T04:06:13Z)
• FAC$^2$E: Better Understanding Large Language Model Capabilities by Dissociating Language and Cognition [56.76951887823882]
  Large language models (LLMs) are primarily evaluated by overall performance on various text understanding and generation tasks. We present FAC$2$E, a framework for Fine-grAined and Cognition-grounded LLMs' Capability Evaluation.
  arXiv  Detail & Related papers  (2024-02-29T21:05:37Z)
• Learning to Generate Explainable Stock Predictions using Self-Reflective Large Language Models [54.21695754082441]
  We propose a framework to teach Large Language Models (LLMs) to generate explainable stock predictions. A reflective agent learns how to explain past stock movements through self-reasoning, while the PPO trainer trains the model to generate the most likely explanations. Our framework can outperform both traditional deep-learning and LLM methods in prediction accuracy and Matthews correlation coefficient.
  arXiv  Detail & Related papers  (2024-02-06T03:18:58Z)
• Enabling Large Language Models to Learn from Rules [99.16680531261987]
  We are inspired that humans can learn the new tasks or knowledge in another way by learning from rules. We propose rule distillation, which first uses the strong in-context abilities of LLMs to extract the knowledge from the textual rules. Our experiments show that making LLMs learn from rules by our method is much more efficient than example-based learning in both the sample size and generalization ability.
  arXiv  Detail & Related papers  (2023-11-15T11:42:41Z)
• Event knowledge in large language models: the gap between the impossible and the unlikely [46.540380831486125]
  We show that pre-trained large language models (LLMs) possess substantial event knowledge. They almost always assign higher likelihood to possible vs. impossible events. However, they show less consistent preferences for likely vs. unlikely events.
  arXiv  Detail & Related papers  (2022-12-02T23:43:18Z)
• The Goldilocks of Pragmatic Understanding: Fine-Tuning Strategy Matters for Implicature Resolution by LLMs [26.118193748582197]
  We evaluate four categories of widely used state-of-the-art models. We find that, despite only evaluating on utterances that require a binary inference, models in three of these categories perform close to random. These results suggest that certain fine-tuning strategies are far better at inducing pragmatic understanding in models.
  arXiv  Detail & Related papers  (2022-10-26T19:04:23Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.