Piotr Borycki and Kornel Howil at PRE-NeurIPS 2025
Table of Contents
Piotr Borycki and Kornel Howil at PRE-NeurIPS 2025
Piotr Borycki and Kornel Howil participated in PRE-NeurIPS 2025, actively representing the group and presenting the latest research achievements of the team.
Piotr works on interpretability of neural networks and explainability methods for machine-learning models. During the event, he presented results from the EPIC project, which introduces a novel prototype-based approach to explaining decisions of pretrained image classifiers — a paper accepted to AAAI. Piotr is also a co-author of CLIPGaussian, a framework enabling universal, multimodal style transfer in 3D using Gaussian Splatting — a NeurIPS-accepted contribution and one of the key research outputs of the project and grant.
Kornel specializes in generative models and neural 3D rendering, with a strong focus on practical and scalable applications of Gaussian Splatting. He was one of the main authors and speakers of CLIPGaussian, which he presented with the team at PRE-NeurIPS 2025, also representing the group in dissemination activities published by ELLIS Unit Warsaw on LinkedIn. His research combines multimodal modeling (e.g., CLIP-type models) with 3D style transformations, producing solutions that are valuable both scientifically and for creative AI applications.
Both researchers continue to actively represent the team internationally — in December at NeurIPS in San Diego (USA), followed by AAAI in Singapore in January — strengthening the global visibility of the project and advancing its contribution to research in 3D artificial intelligence and model interpretability.
