Saved in:
Bibliographic Details
Main Authors: Zhu, Yihua, Liu, Qianying, Wang, Jiaxin, Cheng, Fei, Liu, Chaoran, Aizawa, Akiko, Kurohashi, Sadao, Shimodaira, Hidetoshi
Format: Preprint
Published: 2026
Subjects:
Online Access:https://arxiv.org/abs/2601.02931
Tags: Add Tag
No Tags, Be the first to tag this record!
Table of Contents:
  • Autoregressive LLMs perform well on relational tasks that require linking entities via relational words (e.g., father/son, friend), but it is unclear whether they learn the logical semantics of such relations (e.g., symmetry and inversion logic) and, if so, whether reversal-type failures arise from missing relational semantics or left-to-right order bias. We propose a controlled Knowledge Graph-based synthetic framework that generates text from symmetric/inverse triples, train GPT-style autoregressive models from scratch, and evaluate memorization, logical inference, and in-context generalization to unseen entities to address these questions. We find a sharp phase transition in which relational semantics emerge with sufficient logic-bearing supervision, even in shallow (2-3 layer) models, and that successful generalization aligns with stable intermediate-layer signals. Finally, order-matched forward/reverse tests and a diffusion baseline indicate that reversal failures are primarily driven by autoregressive order bias rather than deficient inversion semantics.