Saved in:
Bibliographic Details
Main Authors: Xia, Wei, Tang, Haowen, Li, Luozheng
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2601.04207
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866911359971622912
author Xia, Wei
Tang, Haowen
Li, Luozheng
author_facet Xia, Wei
Tang, Haowen
Li, Luozheng
contents LLMs internally organize political ideology along low-dimensional structures that are partially, but not fully aligned with human ideological space. This misalignment is systematic, model specific, and measurable. We introduce a lightweight linear probe that both quantifies the misalignment and minimally corrects the output layer. This paper introduces a simple and efficient method for aligning models with specific user opinions. Instead of retraining the model, we calculated a bias score from its internal features and directly adjusted the final output probabilities. This solution is practical and low-cost and preserves the original reasoning power of the model.
format Preprint
id arxiv_https___arxiv_org_abs_2601_04207
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Ideology as a Problem: Lightweight Logit Steering for Annotator-Specific Alignment in Social Media Analysis
Xia, Wei
Tang, Haowen
Li, Luozheng
Computation and Language
Artificial Intelligence
Social and Information Networks
I.2.7; K.4.1
LLMs internally organize political ideology along low-dimensional structures that are partially, but not fully aligned with human ideological space. This misalignment is systematic, model specific, and measurable. We introduce a lightweight linear probe that both quantifies the misalignment and minimally corrects the output layer. This paper introduces a simple and efficient method for aligning models with specific user opinions. Instead of retraining the model, we calculated a bias score from its internal features and directly adjusted the final output probabilities. This solution is practical and low-cost and preserves the original reasoning power of the model.
title Ideology as a Problem: Lightweight Logit Steering for Annotator-Specific Alignment in Social Media Analysis
topic Computation and Language
Artificial Intelligence
Social and Information Networks
I.2.7; K.4.1
url https://arxiv.org/abs/2601.04207