Our researchers have been involved in some important research papers as part of their hub activities. Here is a selection of this research.

2025 / 2026

Title Hub researchers listed as authors View paper
A Diffusive Classification Loss for Learning Energy-based Generative Models José Miguel Hernández‑Lobato arXiv
CCDb+: Enhanced Annotations and Multi-Modal Benchmark for Natural Dyadic Conversations Yukun Lai ACM Multimedia
CREPE: Controlling Diffusion with Replica Exchange José Miguel Hernández‑Lobato arXiv
DiffRatio: Training One-Step Diffusion Models Without Teacher Supervision Mingtian Zhang; José Miguel Hernández‑Lobato; David Barber arXiv
Domain-Adapted Diffusion Model for PROTAC Linker Design Through the Lens of Density Ratio Zixing Song; José Miguel Hernández‑Lobato OpenReview
Efficient Deconvolution in Populational Inverse Problems Mark Girolami arXiv
EnvSDD: Benchmarking Environmental Sound Deepfake Detection Mark D. Plumbley arXiv
FedEDM: Federated Equivariant Diffusion Model for 3D Molecule Generation with Enhanced Communication Efficiency Zixing Song; José Miguel Hernández‑Lobato ACM
LGA-Net: Learning Local and Global Affinities for Sparse Scribble-based Image Colorization Yukun Lai ICCV / CVF ICCV
Neural Mutual Information Estimation with Vector Copulas Michael U. Gutmann arXiv
RESCUE: Crowd Evacuation Simulation via Controlling SDM-United Characters Yukun Lai ICCV / CVF ICCV
Single-Image 3D Human Reconstruction with 3D-Aware Diffusion Priors and Facial Enhancement Yukun Lai SIGGRAPH Asia
Softly Constrained Denoisers for Diffusion Models Samuel Kaski; Mingfei Sun OpenReview
Summary of The Inaugural Music Source Restoration Challenge Mark D. Plumbley arXiv
Towards Reliable Objective Evaluation Metrics for Generative Singing Voice Separation Models Mark D. Plumbley IEEE Xplore
Track and Tweak: Monitoring and Improving Group Fairness for Temporal Graph Neural Networks in Real Time Zixing Song; José Miguel Hernández‑Lobato ACM KDD
BiBBDM: Bidirectional Image Translation With Brownian Bridge Diffusion Models Yukun Lai IEEE TPAMI

2024/25

Title Hub researchers listed as authors View paper
A Primer on Variational Inference for Physics-Informed Deep Generative Modelling Mark Girolami Royal Society
AudioSetCaps: An Enriched Audio-Caption Dataset Using Automated Generation Pipeline With Large Audio and Language Models Mark Plumbley IEEE Xplore
Efficient Prior Calibration from Indirect Data Mark Girolami SIAM Journal on Scientific Computing
Training Neural Samplers with Reverse Diffusive KL Divergence Mingtian Zhang, David Barber, José Miguel Hernández‑Lobato arXiv
Your Finetuned Large Language Model is Already a Powerful Out-of-distribution Detector Andi Zhang arXiv
Efficient Deconvolution in Populational Inverse Problems Mark Girolami arXiv