Synapse
⌘+K
Synapse
PulseExploreClubsResearchersJournals
Instagram
HomeClubsExplore
October 12, 2025Open Access

Benchmarking Gender and Political Bias in Large Language Models

View Full Paper
Ask AI
Bookmark
Share

Authors

JYJinrui YangXHXudong HanTBTimothy Baldwin

Discussion

Loading...

Member takes

Overview

Benchmark assesses gender classification and vote prediction in large language models, highlighting bias.

Key Points

  • Large language models frequently misclassify female Members of the European Parliament as male, indicating a systematic bias.
  • Evaluation shows that LLMs tend to favor centrist political groups while exhibiting reduced accuracy on far-left and far-right categories.
  • Proprietary models like GPT-4o outperform open-weight alternatives in robustness, fairness, and accuracy on politically sensitive tasks.
  • The EuroParlVote dataset provides essential data for future research on fairness and accountability in natural language processing.

Cite This Study

Yang et al. (2025) studied this question.

synapsesocial.com/papers/68ebffcfdef9fcb308ff2666https://doi.org/10.48550/arxiv.2509.06164
View Full Paper
Ask AI
Bookmark
Share

Also Consider

Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context:

  1. 1Gender and Political Bias in Large Language Models: A Demonstration Platform2025
  2. 2Assessing Political Bias in Large Language Models2024 · 3 citations
  3. 3Llama meets EU: Investigating the European Political Spectrum through the Lens of LLMs2024
  4. 4Measuring Political Bias in Large Language Models: What Is Said and How It Is Said2024 · 3 citations
  5. 5Investigating LLMs as Voting Assistants via Contextual Augmentation: A Case Study on the European Parliament Elections 20242024