IndiCASA: A Dataset and Bias Evaluation Framework in LLMs Using Contrastive Embedding Similarity in the Indian Context
Preprint
Abstract
Large Language Models (LLMs) have gained significant traction across critical domains owing to their impressive contextual understanding and generative capabilities. However, their increasing deployment in high stakes applications necessitates rigorous evaluation of embedded biases, particularly in culturally diverse contexts like India where existing embedding-based bias assessment methods often fall short in capturing nuanced stereotypes. We propose an evaluation framework based on a encoder trained using contrastive learning that captures fine-grained bias through embedding similarity. We also introduce a novel dataset - IndiCASA (IndiBias-based Contextually Aligned Stereotypes and Anti-stereotypes) comprising 2,575 human-validated sentences spanning five demographic axes: caste, gender, religion, disability, and socioeconomic status. Our evaluation of multiple open-weight LLMs reveals that all models exhibit some degree of stereotypical bias, with disability related biases being notably persistent, and religion bias generally lower likely due to global debiasing efforts demonstrating the need for fairer model development.Keywords
All Related Versions
This publication has 20 references indexed in Scilit:
- LExT: Towards Evaluating Trustworthiness of Natural Language Explanations2025
- Human-in-the-loop or AI-in-the-loop? Automate or Collaborate?Proceedings of the AAAI Conference on Artificial Intelligence, 2025
- Detecting implicit biases of large language models with Bayesian hypothesis testingScientific Reports, 2025
- VERB: Visualizing and Interpreting Bias Mitigation Techniques Geometrically for Word RepresentationsACM Transactions on Interactive Intelligent Systems, 2024
- On the Dangers of Stochastic Parrots2021
- Social Bias Frames: Reasoning about Social and Power Implications of Language2020
- Social Biases in NLP Models as Barriers for Persons with Disabilities2020
- Deep learning for chest radiograph diagnosis: A retrospective comparison of the CheXNeXt algorithm to practicing radiologistsPLOS Medicine, 2018
- Semantics derived automatically from language corpora contain human-like biasesScience, 2017
- Dimensionality Reduction by Learning an Invariant Mapping2006