Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Tikhomirova, Taisiia, Wulff, Dirk U.
Format: Preprint
Veröffentlicht: 2026
Schlagworte:
Online-Zugang:https://arxiv.org/abs/2601.03798
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866912806855507968
author Tikhomirova, Taisiia
Wulff, Dirk U.
author_facet Tikhomirova, Taisiia
Wulff, Dirk U.
contents Understanding where transformer language models encode psychologically meaningful aspects of meaning is essential for both theory and practice. We conduct a systematic layer-wise probing study of 58 psycholinguistic features across 10 transformer models, spanning encoder-only and decoder-only architectures, and compare three embedding extraction methods. We find that apparent localization of meaning is strongly method-dependent: contextualized embeddings yield higher feature-specific selectivity and different layer-wise profiles than isolated embeddings. Across models and methods, final-layer representations are rarely optimal for recovering psycholinguistic information with linear probes. Despite these differences, models exhibit a shared depth ordering of meaning dimensions, with lexical properties peaking earlier and experiential and affective dimensions peaking later. Together, these results show that where meaning "lives" in transformer models reflects an interaction between methodological choices and architectural constraints.
format Preprint
id arxiv_https___arxiv_org_abs_2601_03798
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Where meaning lives: Layer-wise accessibility of psycholinguistic features in encoder and decoder language models
Tikhomirova, Taisiia
Wulff, Dirk U.
Computation and Language
Artificial Intelligence
Understanding where transformer language models encode psychologically meaningful aspects of meaning is essential for both theory and practice. We conduct a systematic layer-wise probing study of 58 psycholinguistic features across 10 transformer models, spanning encoder-only and decoder-only architectures, and compare three embedding extraction methods. We find that apparent localization of meaning is strongly method-dependent: contextualized embeddings yield higher feature-specific selectivity and different layer-wise profiles than isolated embeddings. Across models and methods, final-layer representations are rarely optimal for recovering psycholinguistic information with linear probes. Despite these differences, models exhibit a shared depth ordering of meaning dimensions, with lexical properties peaking earlier and experiential and affective dimensions peaking later. Together, these results show that where meaning "lives" in transformer models reflects an interaction between methodological choices and architectural constraints.
title Where meaning lives: Layer-wise accessibility of psycholinguistic features in encoder and decoder language models
topic Computation and Language
Artificial Intelligence
url https://arxiv.org/abs/2601.03798