We invite participation to the 4th shared task on interactive (personalized) knowledge assistance (iKAT 2026), which will be hosted as part of the 10th Workshop on Search-Oriented Conversational AI (SCAI)https://scai.info/scai-2026/ , colocated with SIGIR 2026 in Melbourne, Australia.
* Name: Interactive Knowledge Assistance Track (iKAT) * Organizers: Marcel Gohsenhttps://www.uni-weimar.de/en/media/chairs/computer-science-department/webis/people/gohsen/, Mohammad Aliannejadihttps://aliannejadi.com/, Simon Luparthttps://simonlupart.github.io/, Zahra Abbasiantaebhttps://zahraabbasiantaeb.github.io/, Nailia Mirzakhmedovahttps://www.uni-weimar.de/en/media/chairs/computer-science-department/webis/people/#mirzakhmedova, Johannes Kieselhttps://kiesels.de/johannes, Jeff Daltonhttps://www.dcs.gla.ac.uk/~jeff/ * Preliminary timeline (exact dates might still be subject of change): * Registration deadline: May 27, 2026 * System submission opens: May 27, 2026 * System submission deadline: June 10, 2026 * Report submission deadline: July 17, 2026 * SCAI Workshop: July 24, 2026 * Website: https://www.trecikat.com/ * Registration: https://form.jotform.com/260753949743066
Motivation
While the conversational interaction paradigm becomes more and more prevalent for search, the challenge of adapting system responses to individual users (i.e., personalization) still remains a challenge for conversational retrieval systems. Since answer relevance is highly subjective, personalization is necessary to build effective user-centric retrieval systems. From the beginning, personalization of conversational search is the central goal of the Interactive Knowledge Assistance Track (iKAT). The iKAT shared task can look back on four years under former name, Conversational Assistance Track (CAsT), and three years under the current name, iKAT, hosted at TREC. In 2026, iKAT will be organized in collaboration with the SCAI workshop that will take place at SIGIR 2026 in Melbourne, Australia.
Task Overview
Participants of iKAT will receive the following resources to provide their run submissions.
1. Test dataset that contains the following information: a) Topics b) Example conversations c) Personal Text Knowledge Bases (PTKB) of the users 2. Passage collection (iKAT subset of ClueWeb22-B) 3. Access to the user simulation API (Sim.APIhttps://github.com/marcel-gohsen/user-simulation-api) and documentation.
Participants submit conversational search systems that, for each real-time simulated user turn, retrieves and ranks relevant passages from the passage collection. Then, the system use the ranked passages to generate a personalized response based on the user’s PTKB, utterance, and context. This response will be passed back to the simulated user, which again produces an utterance. This process will be repeated until the user terminates the conversation. The systems have to provide the generated text response, the ranked list of passages (citations), and the PTKB statements that were deemed relevant for each produced turn.
Submission
Participants submit their runs via an API and submit a report to describe their approaches. Every participating team will be invited to present their approaches during the SCAI workshop and the reports will be released as part of our proceedings.
Further Details
For more information and updates on the task please keep an eye on our guidelines and on our iKAT websitehttps://www.trecikat.com/.