Negar Arabzadeh

Information retrieval researcher working on evaluation methodology, and in particular on how LLM-based relevance judgments behave when they are used in place of human assessors.

Known For

• Benchmarking LLM-based Relevance Judgment Methods (arXiv:2504.12558, 2025) with Charles L. A. Clarke — a side-by-side benchmark of binary, graded, pairwise-preference and nugget-based LLM judgments, showing that agreement with human labels and agreement with human system rankings are maximized by different methods

Articles

• Benchmarking LLM-based Relevance Judgment Methods

Related Concepts

• LLM as Judge
• Search Evaluation
• Pairwise Relevance Evaluation

People

• Charles L. A. Clarke — co-author