Evaluation of LLMs-based Hidden States as Author Representations for Psychological Human-Centered NLP Tasks

Fuente: arXiv
Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Soni, Nikita, Chitale, Pranav, Singh, Khushboo, Balasubramanian, Niranjan, Schwartz, H. Andrew
Format: Preprint
Veröffentlicht: 2025
Schlagworte:
Online-Zugang:
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866912490142564352
author Soni, Nikita
Chitale, Pranav
Singh, Khushboo
Balasubramanian, Niranjan
Schwartz, H. Andrew
author_facet Soni, Nikita
Chitale, Pranav
Singh, Khushboo
Balasubramanian, Niranjan
Schwartz, H. Andrew
contents Like most of NLP, models for human-centered NLP tasks -- tasks attempting to assess author-level information -- predominantly use representations derived from hidden states of Transformer-based LLMs. However, what component of the LM is used for the representation varies widely. Moreover, there is a need for Human Language Models (HuLMs) that implicitly model the author and provide a user-level hidden state. Here, we systematically evaluate different ways of representing documents and users using different LM and HuLM architectures to predict task outcomes as both dynamically changing states and averaged trait-like user-level attributes of valence, arousal, empathy, and distress. We find that representing documents as an average of the token hidden states performs the best generally. Further, while a user-level hidden state itself is rarely the best representation, we find its inclusion in the model strengthens token or document embeddings used to derive document- and user-level representations resulting in best performances.
format Preprint
id arxiv_https___arxiv_org_abs_2503_00124
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Evaluation of LLMs-based Hidden States as Author Representations for Psychological Human-Centered NLP Tasks
Soni, Nikita
Chitale, Pranav
Singh, Khushboo
Balasubramanian, Niranjan
Schwartz, H. Andrew
Computation and Language
Artificial Intelligence
Human-Computer Interaction
Machine Learning
Like most of NLP, models for human-centered NLP tasks -- tasks attempting to assess author-level information -- predominantly use representations derived from hidden states of Transformer-based LLMs. However, what component of the LM is used for the representation varies widely. Moreover, there is a need for Human Language Models (HuLMs) that implicitly model the author and provide a user-level hidden state. Here, we systematically evaluate different ways of representing documents and users using different LM and HuLM architectures to predict task outcomes as both dynamically changing states and averaged trait-like user-level attributes of valence, arousal, empathy, and distress. We find that representing documents as an average of the token hidden states performs the best generally. Further, while a user-level hidden state itself is rarely the best representation, we find its inclusion in the model strengthens token or document embeddings used to derive document- and user-level representations resulting in best performances.
title Evaluation of LLMs-based Hidden States as Author Representations for Psychological Human-Centered NLP Tasks
topic Computation and Language
Artificial Intelligence
Human-Computer Interaction
Machine Learning
url https://arxiv.org/abs/2503.00124