7 papers · 1 filter
Unbalanced W-Flow: Unbalanced Wasserstein Drifting under Finite-Batch Marginal Fluctuations
Jongwook Jeon, Mingue Park, Jiwon Choi +3
Teacher-free one-step generative models such as W-Flow learn directly from distributional objectives, with the Sinkhorn divergence providing the strongest empirical performance amo…
I-SplineFlow: Learning Monotone Spline Stochastic Interpolant Schedulers for Few-Step Generation
Md Sakib Hossain Shovon, Md Rifat Ur Rahman, Md Abtahi Majeed Chowdhury +3
Few-step generation with pretrained diffusion and flow models can be accelerated by lightweight training that optimizes the sampling trajectory rather than the network. A recent ap…
NoiseTilt: Noise-Tilted Reverse Kernels for Diffusion Reward Alignment
Jisung Hwang, Yunhong Min, Jaihoon Kim +2
We introduce the Noise-Tilted Reverse Kernel (NTRK), a reward-guided diffusion sampler that injects reward gradients through the noise term, leaving the pretrained reverse kernel u…
MUNI: Multimodal Unified Latent Diffusion for Coherent Any-to-Any Generation
Kyeongmin Yeo, Yunhong Min, Minhyuk Sung
We introduce MUNI, an end-to-end multimodal latent diffusion framework for any-to-any generation that unifies subset-conditioned cross-modal generation and unconditional joint samp…
Drifting Field Policy: A One-Step Generative Policy via Wasserstein Gradient Flow
Juil Koo, Mingue Park, Jiwon Choi +2
We propose Drifting Field Policy (DFP), a non-ODE one-step generative policy built on the drifting model paradigm. We frame the policy update as a reverse-KL Wasserstein-2 gradient…
BézierFlow: Learning Bézier Stochastic Interpolant Schedulers for Few-Step Generation
Yunhong Min, Juil Koo, Seungwoo Yoo +1
We introduce BézierFlow, a lightweight training approach for few-step generation with pretrained diffusion and flow models. BézierFlow achieves a 2-3x performance improvement for s…
Psi-Sampler: Initial Particle Sampling for SMC-Based Inference-Time Reward Alignment in Score Models
Taehoon Yoon, Yunhong Min, Kyeongmin Yeo +1
We introduce -Sampler, an SMC-based framework incorporating pCNL-based initial particle sampling for effective inference-time reward alignment with a score-based generative mode…