LongEval CLEF 2025 Lab
Longitudinal Evaluation of Model Performance
| Description | Dates | Organizers | Tasks | Data | Submissions | 2024 | 2023 |
We have released the Training Collections for the
WebRetrieval and
SciRetrieval Tasks!
Registration is open here
April 2025: Test data release
Until 2024, the information retrieval tasks of LongEval did focus only on Web documents retrieval. In 2025, we enlarge the scope of LongEval as we want to study the potential differences, if any, between two retrieval contexts. The Web retrieval context is a classical Web case, in which very short queries are asked and the very top documents are considered. The scientific search contains potentially longer queries, and users are looking deeper in the result lists.
Both LongEval tasks use a sequence of datasets collected at different points in time. The (time) distances between two datasets are called “lags.” The IR systems that participants design are evaluated on the different lags, computing the differences in evaluation metrics between lags.
The CLEF 2025 LongEval Lab encourages participants to develop temporal information retrieval systems that survive through dynamic temporal text changes, introducing time as a new dimension for ranking models performance.
LongEval organizers: longeval-ir-task@univ-grenoble-alpes.fr
Join our slack channel for any question.