academic-recommendation

This paper investigates authority bias in LLMs used as conversational search engines for academic paper recommendation, finding that LLMs show significant preference for papers based on authority signals like author prestige and citations, with debiasing only partially effective.