Selected work

Representative Papers

Entering the Era of Discrete Diffusion Models: A Benchmark for Schrödinger Bridges and Entropic Optimal Transport

Sep 27, 2025

Existing discrete diffusion and flow models lack reliable evaluation methods for solving entropy-regularized optimal transport (EOT) and Schrödinger bridge (SB) problems. Method: We introduce the first discrete SB benchmark framework with closed-form analytical solutions, enabling precise and reproducible validation of discrete diffusion models. We propose two efficient algorithms—DLightSB and DLightSB-M—and extend them to α-CSBM, enabling the first systematic evaluation of discrete-domain SB methods. Leveraging EOT theory and dynamic SB solvers, we construct multiple high-dimensional analytically tractable probability distribution pairs and conduct comprehensive performance comparisons of both classical and novel solvers under unified experimental settings. Contribution/Results: This work fills a critical gap in interpretable evaluation for discrete generative models and establishes essential infrastructure for grounding SB theory in practice. Our benchmark and algorithms facilitate rigorous, transparent, and reproducible assessment of discrete SB-based generative modeling, advancing both theoretical understanding and empirical development.

0 citationsRead paper

Universal Inverse Distillation for Matching Models with Real-Data Supervision (No GANs)

Sep 26, 2025

Generative models—including diffusion models, flow matching, and related frameworks—suffer from slow inference. Existing knowledge distillation methods are either framework-specific or rely on data-free paradigms; incorporating real data typically necessitates complex adversarial training. Method: We propose the first general-purpose one-step distillation framework, unifying diverse matching-based generative models (e.g., diffusion, flow matching, bridge matching, and stochastic interpolation). Grounded in reverse distillation theory, we introduce a trajectory alignment loss that directly integrates real-data supervision—without GANs or discriminators. Contribution/Results: Our method achieves high-fidelity single-step generation across multiple tasks, significantly accelerating inference while preserving cross-model generalizability and stability. It overcomes dual limitations of conventional distillation: dependence on model-specific architectures and restrictive data strategies.

0 citationsRead paper

Recent publications

Latest Papers

Entering the Era of Discrete Diffusion Models: A Benchmark for Schrödinger Bridges and Entropic Optimal Transport

Sep 27, 2025

Existing discrete diffusion and flow models lack reliable evaluation methods for solving entropy-regularized optimal transport (EOT) and Schrödinger bridge (SB) problems. Method: We introduce the first discrete SB benchmark framework with closed-form analytical solutions, enabling precise and reproducible validation of discrete diffusion models. We propose two efficient algorithms—DLightSB and DLightSB-M—and extend them to α-CSBM, enabling the first systematic evaluation of discrete-domain SB methods. Leveraging EOT theory and dynamic SB solvers, we construct multiple high-dimensional analytically tractable probability distribution pairs and conduct comprehensive performance comparisons of both classical and novel solvers under unified experimental settings. Contribution/Results: This work fills a critical gap in interpretable evaluation for discrete generative models and establishes essential infrastructure for grounding SB theory in practice. Our benchmark and algorithms facilitate rigorous, transparent, and reproducible assessment of discrete SB-based generative modeling, advancing both theoretical understanding and empirical development.

0 citationsRead paper

Universal Inverse Distillation for Matching Models with Real-Data Supervision (No GANs)

Sep 26, 2025

Generative models—including diffusion models, flow matching, and related frameworks—suffer from slow inference. Existing knowledge distillation methods are either framework-specific or rely on data-free paradigms; incorporating real data typically necessitates complex adversarial training. Method: We propose the first general-purpose one-step distillation framework, unifying diverse matching-based generative models (e.g., diffusion, flow matching, bridge matching, and stochastic interpolation). Grounded in reverse distillation theory, we introduce a trajectory alignment loss that directly integrates real-data supervision—without GANs or discriminators. Contribution/Results: Our method achieves high-fidelity single-step generation across multiple tasks, significantly accelerating inference while preserving cross-model generalizability and stability. It overcomes dual limitations of conventional distillation: dependence on model-specific architectures and restrictive data strategies.

0 citationsRead paper