A Conditional Diffusion Approach to Multidimensional Unfolding Across Physics Processes

Towards a universal unfolder: cDDPMs learn P(x|y) conditioned on detector stats, enabling multidimensional, non-iterative unfolding that reduces truth bias.