Lorenzo Olearo

Diffusion models have dramatically advanced text-to-image generation in recent years, translating abstract concepts into high-fidelity images with remarkable ease. In this work, we examine whether they can also blend distinct concepts, ranging from concrete objects to intangible ideas, into coherent new visual entities under a zero-shot framework. Specifically, concept blending merges the key attributes of multiple concepts (expressed as textual prompts) into a single, novel image that captures the essence of each concept. We investigate four blending methods, each exploiting different aspects of the diffusion pipeline (e.g., prompt scheduling, embedding interpolation, or layer-wise conditioning). Through systematic experimentation across diverse concept categories, such as merging concrete concepts, synthesizing compound words, transferring artistic styles, and blending architectural landmarks, we show that modern diffusion models indeed exhibit creative blending capabilities without further training or fine-tuning. Our extensive user study, involving 100 participants, reveals that no single approach dominates in all scenarios: each blending technique excels under certain conditions, with factors like prompt ordering, conceptual distance, and random seed affecting the outcome. These findings highlight the remarkable compositional potential of diffusion models while exposing their sensitivity to seemingly minor input variations.

Speaker verification is the task of examining a speech signal to authenticate the claimed identity of a speaker as true or false. In order to deal with utterances having different lengths, and to accumulate information along the time dimension, different temporal aggregators have been proposed inside speaker verification pipelines. In this paper we investigate the behavior of five different temporal aggregators in the state of art, namely Temporal Average Pooling (TAP), Global Statistical Pooling (GSP), Self-Attentive Pooling (SAP), Attentive Statistical Pooling (ASP), and Vector of Locally Aggregated Descriptors (VLAD) at varying lengths of the two utterances. Starting from a speaker verification method in the state of the art, the experimental results on the VoxCeleb2 dataset show that there is a sweet spot for utterance length where speaker verification performance is higher independently from the temporal aggregator used.