Dissertation
Evaluation of Bias and Robustness in Search and Conversational Systems
Search and conversational systems have become central to how people access information and perform tasks.
- Author
- A Abolghasemi
- Date
- 06 March 2026
- Links
- Thesis in Leiden Repository
With the emergence of large language models (LLMs), information systems have shifted from purely retrieval-based pipelines toward generation and retrieval-augmented generation (RAG). While these advances bring new opportunities, they also introduce challenges such as outdated knowledge, hallucinations, bias, and failures across multi-stage information systems. Ensuring that such systems are robust, unbiased, and trustworthy requires systematic evaluation across a broad range of tasks and contexts.
In this thesis, we investigate how retrieval and generative models behave in nuanced real-world information-seeking scenarios, with a particular focus on robustness and unbiasedness, as essential aspects of building reliable and trustworthy systems.