Mohammad Aliannejadi

I am an assistant professor at the IRLab of the University of Amsterdam, working on conversational and agentic information access. I focus on building and evaluating systems that search and reason over retrieved evidence while interacting with a user in multimodal and mixed-initiative environments. One central question I ask is how to benchmark and evaluate such systems, and to what extent we can trust LLMs-as-judges and user simulators.

Current directions in my work include:

• Conversational and agentic search: retrieval-augmented generation and reasoning over multi-turn conversations, efficient conversational retrieval, and deep research agents.
• Mixed-initiative and multimodal interaction: asking clarifying questions, multimodal clarification, and multimodal RAG systems.
• Benchmarking and evaluation: test collections such as TREC iKAT and Total Recall QA, and the reliability of LLMs-as-judges for relevance and response quality.
• User simulation: simulating users to train and evaluate interactive systems, and infrastructure for running simulation-based shared tasks.
• Query performance prediction, recommendation, and fairness: predicting retrieval quality without judgments, cross-market and next-basket recommendation, and bias in ranking.

I am passionate about building shared resources for the community. I lead the TREC Interactive Knowledge Assistance Track (iKAT), and before that led the mixed-initiative subtask of TREC CAsT, chaired the ClariQ challenge (ConvAI3), and co-organized the IGLU competition at NeurIPS. With my students and collaborators I have released datasets and benchmarks such as Total Recall QA, the iKAT collections, ClariQ, Qulac, ANTIQUE, and XMarket (see data & code).

I am an active member of the IR and NLP community. I was the lead organizer of the European Summer School in Information Retrieval (ESSIR 2024) the Dutch-Belgian IR workshop (DIR 2021), and the XMRec workshop at RecSys 2021; program co-chair of ICTIR 2023, lab co-chair of CLEF 2023, and tutorial co-chair of ECIR 2024. I co-organize workshops including SCAI, LLM4Eval, QPP++, Eval4RAG, and MICROS, have given tutorials on conversational search and query performance prediction at SIGIR, WWW, ECIR, and SIGIR-AP, and served as ACM SIGIR Student Liaison for Europe and Africa. I am an ELLIS Scholar, and in 2025 I received the SIGIR Early Career Researcher Award for Community Engagement. See service & talks for more.

Before joining IRLab as a faculty member, I was a postdoctoral researcher at the University of Amsterdam. I obtained my Ph.D. from the Università della Svizzera italiana (USI) in Lugano, Switzerland, with a thesis on modeling user information needs on mobile devices, including a research visit to CIIR at UMass Amherst. Before that, I completed my M.Sc. at Amirkabir University of Technology (Tehran Polytechnic).

Prospective students: I am always happy to hear from motivated students interested in conversational and agentic search, multimodal RAG, evaluation with LLMs, or user simulation. Open positions are announced on the IRLab website. Current UvA M.Sc. students who want to do a thesis with me can email me with a short description of their interests.