We’ll be presenting eight papers at NeurIPS 2026. Two are in the Main Track, one is in the Datasets & Benchmarks track and five are at workshops. They cover a good part of what we work on: counterfactual explanations, rare-event simulation, film and image restoration, the geometry of learned representations, medical imaging and LLMs in clinical triage. Below is the full list with abstracts. Names in bold are genwro.AI members. ...
What If AI Could Rewrite Its Decisions? In a world increasingly shaped by algorithmic decisions, a fundamental question emerges: when AI makes a decision about us, how can we understand why and — more importantly — what we could change to get a different outcome? Explainable AI (XAI) is all about answering those questions, pulling back the curtain from black-box algorithms. But while most XAI tools tell you what went wrong, one approach — counterfactual explanations (CFEs) — goes further. CFEs don’t just tell you why an AI decided what it did, but shows you how that decision could be different. ...
Hey there! Welcome to the genwro.AI Blog 🚀 We’re thrilled to kick off our research blog! Since you’re here, let us give you the scoop on who we are and what we’re planning to share. Meet genwro.AI We are a research group from Wrocław University of Science and Technology established in 2019. Our research spans several key areas in artificial intelligence and machine learning: generative models: developing and advancing state-of-the-art architectures for content generation and synthesis, 3D representation learning: exploring novel approaches to understanding and representing three-dimensional structures in machine learning systems, few-shot learning: investigating methods to enable effective learning from limited data samples, uncertainty estimation: advancing techniques for reliable uncertainty quantification and out-of-distribution detection in ML models, computer vision: focusing on image segmentation, restoration, and video generation with particular emphasis on real-world applications, explainable AI: developing methods for interpretable machine learning, with emphasis on counterfactual explanations and model transparency, probabilistic models: investigating Bayesian approaches and probabilistic frameworks for robust machine learning systems. We’re not working in isolation — we’re part of a broader research community, collaborating with amazing teams from places like University of Cambridge, Imperial College London, ETH Zurich, and several other fantastic institutions. ...