Query performance prediction (QPP), a.k.a. query difficulty prediction, is a key task in information retrieval (IR).
QPP has been studied for decades in the IR community.
The task of QPP is defined as estimating search effectiveness without human relevance judgments.
We are excited to announce the QPP++ 2025 workshop, titled "Query Performance Prediction and its Applications in the Era of Large Language Models," co-located with the 47th European Conference on Information Retrieval (ECIR 2025), from 6th to 10th April 2025 in Lucca, Italy.
In this workshop, we aim to bring together researchers and practitioners from academia and industry to discuss new perspectives on QPP in the era of large language models (LLMs).
QPP++ 2025 is the continuation of the QPP++ 2023 workshop held in ECIR 2023.
Given the rapid advancement of LLMs, we present examples of topics we encourage authors to explore in this workshop:
Beyond the above aspects, we also welcome submissions on other QPP-related aspects. There is a need for deeper exploration in other QPP-related areas, including, but not limited to, the following: QPP for conversational search, recommendation systems, question answering, and fairness. We aim to stimulate discussions on these topics.
We welcome manuscripts about any QPP-related subjects. We welcome a diverse range of submission types:
Note that submissions be written in English.
For original papers, double submission is not allowed except for papers already on ArXiv.
The papers (.pdf format) should be submitted using the EasyChair submission system at https://easychair.org/conferences/?conf=qpp2025.
The paper template can be found at https://ceur-ws.org/HOWTOSUBMIT.html.
The review process is single-blind. Each manuscript will be peer-reviewed by at least three programme committee (PC) members.
Authors of accepted papers will be invited to give oral presentations at the workshop.
The accepted papers will be published in the CEUR-WS.org proceedings series.
Since the proceedings only accept papers with a minimum of 4 pages, we will combine 1-page abstract submissions into one or more papers for publication.
The best papers submitted to the QPP++ 2025 workshop will be invited for an extension on the TOIS Special Issue on "Query Performance Prediction Towards Novel Information Retrieval Paradigms" . The authors of the best papers are invited to implement suggestions and comments provided by the other participants to the workshop and submit an extended version of their work that will be published on the Transactions on Information Systems (TOIS).
To facilitate workshop discussions, we encourage (though it is not mandatory) authors to use common, publicly available benchmarks. This will help authors share their experimental experiences and insights more effectively. We also plan to set a discussion session for sharing experience on common datasets.
We also recommend open-source implementations of QPP methods:
The Role of Query Performance Prediction in Developing Adaptive Search and RAG Systems
Abstract
Query Performance Prediction (QPP) has mainly been studied for more than two decades as a focused sub-topic in Information Retrieval. The community has worked towards developing more effective models for predicting the performance of a diverse range of ranking models, and in this talk I’m going to provide a very brief review of the existing classes of QPP models from unsupervised to supervised to query-variant based approaches. However, a more important problem to which I want to draw the attention of the research community is that of a downstream application of QPP in developing adaptive systems. I’m going to talk about two distinct use-cases: the first, a more direct application of QPP for improving the workflow of IR systems by incorporating a dynamic query-specific pipeline. And the second, a more subtle connection between QPP and RAG, where I will be first talking about how QPP techniques might be used to estimate the usefulness of a RAG context eventually providing some high-level pointers on how this might actually be applied to develop input-specific adaptive RAG pipelines.
Bio
Debasis Ganguly has been working as an assistant professor in the University of Glasgow, UK. He was a former research scientist in IBM Research, Ireland, where he worked on information extraction and verification of scientific claims. His general research goals are directed towards predicting performance of AI models, fair, interpretable and trustworthy AI models, and privacy-preserving AI models. Over the years he has published over a hundred research papers in top tier conferences and journals, namely SIGIR, CIKM, ECIR, TOIS, IPM etc.
QPP++ 2025 is a half-day workshop (9:00–12:30), which will be held on the last day of ECIR 2025 (10th April 2025) at the IMT School for Advanced Studies Lucca. Below is the detailed schedule.
Time | Duration | Session | Speaker(s) / Notes |
---|---|---|---|
09:00–09:10 | 10 min | Opening and Welcome | — |
09:10–09:25 | 15 min | Corpora Performance Prediction [PDF] |
Andrew Parry, Jan Heinrich Merker, Simon Ruth, Maik Fröbe, Harrisen Scells |
09:25–09:40 | 15 min | Revisiting Query Variants: The Advantage of Retrieval Over Generation of Query Variants for Effective QPP [PDF] |
Fangzheng Tian, Debasis Ganguly, Craig Macdonald |
09:40-09:55 | 15 min | Robust Query Performance Prediction for Dense Retrievers via Adaptive Disturbance Generation [PDF] |
Abbas Saleminezhad, Negar Arabzadeh, Radin Hamidi Rad, Soosan Beheshti, Ebrahim Bagheri |
09:55–10:10 | 15 min | PQPP: A Joint Benchmark for Text-to-Image Prompt and Query Performance Prediction [PDF] |
Eduard Poesina, Adriana Costache, Adrian-Gabriel Chifu, Josiane Mothe, Radu Tudor Ionescu |
10:10–10:25 | 15 min | Estimating Query Performance Through Rich Contextualized Query Representations [PDF] |
Sajad Ebrahimi, Maryam Khodabakhsh, Negar Arabzadeh, Ebrahim Bagheri |
10:25–11:00 | 35 min | Break | — |
11:00–11:30 | 30 min | Keynote: The Role of Query Performance Prediction in developing Adaptive Search and RAG Systems | Debasis Ganguly |
11:30–11:55 | 25 min | Breakout Discussion: Evaluating and Applying QPP in Real-world Systems | Attendees will be divided into small groups for discussion, with each group sharing their outcomes at the end. |
11:55–12:20 | 25 min | Breakout Discussion: Future Directions in QPP | Attendees will be divided into small groups for discussion, with each group sharing their outcomes at the end. |
12:20–12:30 | 10 min | Closing Remarks | — |
We thank the following committee members who contribute to the paper review and selection: