Understanding Wacky Weights: A Dissection of SPLADE's Learned Term Importance

Sun, 19 Jul 2026 00:00:00 +0000

Learned sparse retrieval models such as SPLADE combine the effectiveness of neural architectures with the efficiency of inverted indices. As these models assign weights to terms from a fixed vocabulary, interpretability is often touted as a major benefit of these models. However, the emergence of wacky weights, i.e., expansion terms that appear semantically unrelated to the input, limits interpretability. While prior research has anecdotally observed this phenomenon, there is a lack of systematic understanding regarding their origins, prevalence, and contribution to retrieval effectiveness. In this paper, we reproduce SPLADE-v2 to systematically investigate wacky weights across the SPLADE family of models. We present a comprehensive dissection of wacky weights, providing a formal definition of wackiness based on the lexical utility of expansion terms. Furthermore, we introduce a novel measure to compare the prevalence of these tokens across models with varying vocabularies and sparsity levels. Beyond reproducing the original SPLADE-v2, we train it with various loss functions, datasets, and backbone transformers to isolate the factors contributing to wackiness. Our results show that larger vocabularies are associated with a higher prevalence of wacky tokens, while stricter sparsity regularizers are associated with lower prevalence. Finally, we find that wacky weights are used primarily for in-domain effectiveness rather than out-of-domain generalization.

UP2DATE: Shared Task on Systematic Review Updates

Thu, 26 Mar 2026 00:00:00 +0000

In all areas of science, the volume of literature has become overwhelming, and staying current with the bleeding edge of research is increasingly unmanageable. This problem is particularly acute in medicine, where studies are published at a rate of approximately two per minute, and up-to-date information is crucial for decisionmaking processes. Maintaining the evidence base by updating systematic reviews is a costly and time-consuming endeavour, yet there is comparatively little research that tackles this important problem via computational means. We propose a shared task that addresses this challenge in (biomedical) scholarly information access. Specifically, we propose three tasks involved in updating systematic reviews: determining when a review should be updated, retrieving literature for the review update, and classifying whether each piece of retrieved literature should be included in the review. These are difficult problems that have received limited research attention, and improved solutions could have considerable impact.

Carsten Eickhoff | Harry Scells

Understanding Wacky Weights: A Dissection of SPLADE's Learned Term Importance

UP2DATE: Shared Task on Systematic Review Updates