mlsanityData to 7 Oct 2026

POLAR:A Per-User Association Test in Embedding Space

arXiv 2603.15950 ↗First posted 16 Mar 2026cs.CL cs.CY cs.SIcited by 0

Not on the map. Why?

Abstract (arXiv)

Most intrinsic association probes operate at the word, sentence, or corpus level, obscuring author-level variation. We present POLAR (Per-user On-axis Lexical Association Re-port), a per-user lexical association test that runs in the embedding space of a lightly adapted masked language model. Authors are represented by private deterministic to-kens; POLAR projects these vectors onto curated lexicalaxes and reports standardized effects with permutation p-values and Benjamini--Hochberg control. On a balanced bot--human Twitter benchmark, POLAR cleanly separates LLM-driven bots from organic accounts; on an extremist forum,it quantifies strong alignment with slur lexicons and reveals rightward drift over time. The method is modular to new attribute sets and provides concise, per-author diagnostics for computational social science. All code is publicly avail-able at https://github.com/pedroaugtb/POLAR-A-Per-User-Association-Test-in-Embedding-Space.

Key facts. No citing papers in the map yet (data through 7 Oct 2026). Builds on 21 of its 52 references that are in the map. First posted on arXiv 16 Mar 2026.

Builds on

21 of 52 references.

References · 21

  1. 2021LoRA: Low-Rank Adaptation of Large Language Modelscited by 14,084
  2. 2017Inductive Representation Learning on Large Graphscited by 6,284
  3. 2019Parameter-Efficient Transfer Learning for NLPcited by 4,195
  4. 2021Prefix-Tuning: Optimizing Continuous Prompts for Generationcited by 3,878
  5. 2021The Power of Scale for Parameter-Efficient Prompt Tuningcited by 3,769
  6. 2014DeepWalk: Online Learning of Social Representationscited by 1,543
  7. 2017Neural Collaborative Filteringcited by 1,108
  8. 2016Semantics derived automatically from language corpora contain human-like biasescited by 894
  9. 2019Predicting Dynamic Embedding Trajectory in Temporal Interaction Networkscited by 443
  10. 2019On Measuring Social Biases in Sentence Encoderscited by 397
  11. 2019Mitigating Gender Bias in Natural Language Processing: Literature Reviewcited by 287
  12. 2019Measuring Bias in Contextualized Word Representationscited by 275
9 more
  1. 2018Hate Speech Dataset from a White Supremacy Forumcited by 196
  2. 2023Anatomy of an AI-powered malicious social botnetcited by 35
  3. 2019Social Media-based User Embedding: A Literature Reviewcited by 14
  4. 2020Embeddings-Based Clustering for Target Specific Stances: The Case of a Polarized Turkeycited by 12
  5. 2025UQABench: Evaluating User Embedding for Prompting LLMs in Personalized Question Answeringcited by 9
  6. 2022Learning Stance Embeddings from Signed Social Graphscited by 8
  7. 2021Graph embeddings for Abusive Language Detectioncited by 6
  8. 2021Analysis of Twitter Users' Lifestyle Choices using Joint Embedding Modelcited by 4
  9. 2023Using word embeddings to analyse audience effects and individual differences in parenting Subredditscited by 1

Built on by

Not cited yet.