Ojasvi Gupta
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Ojasvi Gupta has a total of 7 co-authors. Here are the top co-authors by number of shared publications:
- Rajesh Jaiswal
Technological University Dublin, Dublin, Ireland
3 shared publications - Francesco Gargiulo
National Research Council (CNR), Institute for High Performance Computing and Networking (ICAR), Naples, Italy
2 shared publications - Lidia Marassi
University of Naples Federico II
2 shared publications - Stefano Marrone
Department of Electrical Engineering and Information Technology, University of Naples Federico II, Naples, Italy
2 shared publications - Abdelsalam Busalim
Faculty of Computing, Digital and Data, Technological University Dublin, Dublin, Leinster, Ireland
1 shared publication - Keith Quille
Technological University Dublin, Dublin, Ireland
1 shared publication
Latest Publications
- Top Cited PapersOpen Access
Understanding Social Biases in Large Language Models
Background/Objectives: Large Language Models (LLMs) like ChatGPT, LLAMA, and Mistral are widely used for automating tasks such as content creation and data analysis. However, due to their training on publicly available internet data, they may inherit social biases. We aimed to investigate the social biases (i.e., ethnic, gender, and disability biases) in these models and evaluate how different model versions handle them. Methods: We instruction-tuned popular models (like Mistral, LLAMA, and Gemma), and for this we curated a dataset constructed by collecting and modifying diverse data from various public datasets. Prompts were run through a controlled pipeline, and responses were categorized (e.g., biased, confused, repeated, or accurate) and analyzed. Results: We found that models responded differently to bias prompts depending on their version. Fine-tuned models showed fewer overt biases but more confusion or censorship. Disability-related prompts triggered the most consistent biases across models. Conclusions: Bias persists in LLMs despite instruction tuning. Differences between model versions may lead to inconsistent user experiences and hidden harms in downstream applications. Greater transparency and robust fairness testing are essential.
Top Cited PapersOpen AccessUnderstanding Social Biases in Large Language Models
Background/Objectives: Large Language Models (LLMs) like ChatGPT, LLAMA, and Mistral are widely used for automating tasks such as content creation and data analysis. However, due to their training on publicly available internet data, they may inherit social biases. We aimed to investigate the social biases (i.e., ethnic, gender, and disability biases) in these models and evaluate how different model versions handle them. Methods: We instruction-tuned popular models (like Mistral, LLAMA, and Gemma), and for this we curated a dataset constructed by collecting and modifying diverse data from various public datasets. Prompts were run through a controlled pipeline, and responses were categorized (e.g., biased, confused, repeated, or accurate) and analyzed. Results: We found that models responded differently to bias prompts depending on their version. Fine-tuned models showed fewer overt biases but more confusion or censorship. Disability-related prompts triggered the most consistent biases across models. Conclusions: Bias persists in LLMs despite instruction tuning. Differences between model versions may lead to inconsistent user experiences and hidden harms in downstream applications. Greater transparency and robust fairness testing are essential.
- Open Access
- 2 December 2024
- Published by Association for Computing Machinery (ACM)
- p. 8-14
Harmful Prompt Classification for Large Language Models
- 2 December 2024
- Published by Association for Computing Machinery (ACM)
- p. 8-14
Open AccessHarmful Prompt Classification for Large Language Models
- Open Access
- 2 December 2024
- Published by Association for Computing Machinery (ACM)
- p. 63
Understanding Gender and Ethnicity Bias in Large Language Models
- 2 December 2024
- Published by Association for Computing Machinery (ACM)
- p. 63
Open AccessUnderstanding Gender and Ethnicity Bias in Large Language Models