Selected work

Representative Papers

Explaining Process Control Optimisation Recommendations via GradientSHAP and Implicit Differentiation

Jul 16, 2026

This study addresses the challenge of gaining operator trust in industrial process optimization recommendations, which often suffer from insufficient interpretability. The authors propose an efficient attribution method that integrates sensitivity analysis based on the implicit function theorem with GradientSHAP—a novel combination for explaining optimization outputs—and leverages a large language model to generate natural-language explanations tailored for plant operators. Evaluated on a high-pressure grinding roll (HPGR) control optimization task involving 22 input features, the proposed approach achieves a correlation exceeding 0.99 with KernelSHAP attributions while accelerating computation by over 40×, thereby enabling real-time interpretability. The method received positive assessments from domain experts for its clarity and practical utility.

0 citationsRead paper

Fine-Grained Erasure in Text-to-Image Diffusion-based Foundation Models

Mar 25, 2025

Existing concept forgetting methods for text-to-image diffusion models often overlook semantic proximity, leading to collateral damage to semantically related concepts during target concept erasure—the adjacency challenge. This paper introduces FADE, the first adjacency-aware fine-grained forgetting framework. FADE constructs concept neighborhoods via a semantic graph and jointly optimizes three objectives—Expungement, Adjacency preservation, and Guidance fidelity—using its novel Mesh module. It employs gradient-constrained optimization over diffusion model parameters and a multi-objective weighted loss design. Evaluated on six benchmarks including Stanford Dogs, FADE significantly reduces adjacent-concept degradation, improves knowledge retention by ≥12% over state-of-the-art methods, and maintains both thorough concept erasure and model generalization stability.

0 citationsRead paper

Recent publications

Latest Papers

Explaining Process Control Optimisation Recommendations via GradientSHAP and Implicit Differentiation

Jul 16, 2026

This study addresses the challenge of gaining operator trust in industrial process optimization recommendations, which often suffer from insufficient interpretability. The authors propose an efficient attribution method that integrates sensitivity analysis based on the implicit function theorem with GradientSHAP—a novel combination for explaining optimization outputs—and leverages a large language model to generate natural-language explanations tailored for plant operators. Evaluated on a high-pressure grinding roll (HPGR) control optimization task involving 22 input features, the proposed approach achieves a correlation exceeding 0.99 with KernelSHAP attributions while accelerating computation by over 40×, thereby enabling real-time interpretability. The method received positive assessments from domain experts for its clarity and practical utility.

0 citationsRead paper

Fine-Grained Erasure in Text-to-Image Diffusion-based Foundation Models

Mar 25, 2025

Existing concept forgetting methods for text-to-image diffusion models often overlook semantic proximity, leading to collateral damage to semantically related concepts during target concept erasure—the adjacency challenge. This paper introduces FADE, the first adjacency-aware fine-grained forgetting framework. FADE constructs concept neighborhoods via a semantic graph and jointly optimizes three objectives—Expungement, Adjacency preservation, and Guidance fidelity—using its novel Mesh module. It employs gradient-constrained optimization over diffusion model parameters and a multi-objective weighted loss design. Evaluated on six benchmarks including Stanford Dogs, FADE significantly reduces adjacent-concept degradation, improves knowledge retention by ≥12% over state-of-the-art methods, and maintains both thorough concept erasure and model generalization stability.

0 citationsRead paper