ReNeuIR at SIGIR 2024: The Third Workshop on Reaching Efficiency in Neural Information Retrieval


The Information Retrieval (IR) community has a rich history of empirically measuring novel retrieval methods in terms of effectiveness and efficiency. However, as the search ecosystem is developing rapidly, comparatively little attention has been paid to evaluating efficiency in recent years, which raises the question of the cost-benefit ratio between effectiveness and efficiency. In this regard, it has become difficult to compare and contrast systems in an empirically fair way. Factors including hardware configurations, software versioning, experimental settings, and measurement methods all contribute to the difficulty of meaningfully comparing search systems, especially where efficiency is a key component of the evaluation. Furthermore, efficiency is no longer limited to time and space but has found new, challenging dimensions that stretch to resource, sample, and energy efficiency and have implications for users, researchers, and the environment. Examining algorithms and models through the lens of efficiency and its trade-off with effectiveness requires revisiting and establishing new standards and principles, from defining relevant concepts, to designing measures, to creating guidelines for making sense of the significance of findings. The third iteration of ReNeuIR aims to bring the community together to debate these questions and collaboratively test and improve a benchmarking framework for efficiency derived from the discussions of the first two iterations of this workshop. We provide a first prototype of this framework by organizing a shared task track focused on comparability and reproducibility at the workshop.

Proceedings of the 47th International ACM Conference on Research and Development in Information Retrieval (SIGIR 2024)