Randomized trial evaluates text semantic similarity using three embedding aggregation methods, suggesting improved model robustness.
The paper explores methods for aggregating the results of three embedding models to estimate semantic similarity of texts: cosine similarities averaging, vector concatenation, and selection of one of the three cosine similarities based on principal component analysis and singular value decomposition methods. Statistical analysis showed that averaging cosine proximity as an aggregation method shows the best results. The obtained results can be used to improve the robustness of models for semantic proximity evaluation of texts when using embedding ensembles.
No takes yet. Share an insight, caveat, or question.
Ilina et al. (2026) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: