Abstract:When we read, we make predictions about upcoming words; these predictions influence our reading behavior. The success of large language models (LLMs), which, like humans, make predictions about upcoming words, has motivated their use as models of human linguistic prediction. Surprisingly, in the last few years, as LLMs' ability to predict the next word has improved, their ability to explain reading behavior has declined. We argue this is because current LLMs can predict upcoming words much better than human readers can. This 'superhumanness' is driven by LLMs' extensive training data, stronger long-term memory of training examples, and stronger short-term memory. We advocate for LLMs with human-like memory and for new experiments to measure the alignment between humans and LLMs, and outline directions towards achieving these goals.
| Comments: | Accepted to Trends in Cognitive Sciences |
| Subjects: | Computation and Language (cs.CL) |
| Cite as: | arXiv:2510.05141 [cs.CL] |
| (or arXiv:2510.05141v2 [cs.CL] for this version) | |
| https://doi.org/10.48550/arXiv.2510.05141 arXiv-issued DOI via DataCite |
From: Byung-Doh Oh [view email]
[v1]
Wed, 1 Oct 2025 21:53:42 UTC (122 KB)
[v2]
Tue, 26 May 2026 14:52:39 UTC (317 KB)
此内容由惯性聚合(RSS阅读器)自动聚合整理，仅供阅读参考。 原文来自 — 版权归原作者所有。