Selected work

Representative Papers

LogicCBMs: Logic-Enhanced Concept-Based Learning

Dec 08, 2025

Existing concept bottleneck models (CBMs) rely on linear concept combinations, limiting expressivity and hindering modeling of complex semantic relationships. This paper proposes Logic-CBM, the first differentiable logic-driven CBM that embeds propositional logical operations (e.g., AND, OR, NOT) into an end-to-end trainable framework. Leveraging a neuro-symbolic design—integrating differentiable logic gates, gradient straight-through estimators, and joint optimization—it enables nonlinear logical reasoning over concepts. Evaluated on multiple benchmark and synthetic datasets, Logic-CBM achieves significant improvements in prediction accuracy, supports precise concept-level interventions, and preserves strong interpretability and model transparency. Its core contributions are threefold: (1) introducing the first differentiable logic-based concept learning paradigm; (2) overcoming the fundamental limitations of linear concept composition; and (3) unifying the structural rigor of symbolic logic with the learnability of deep neural networks.

0 citationsRead paper

Semiparametric Learning from Open-Set Label Shift Data

Sep 17, 2025

This paper addresses the open-set label shift problem, where the test distribution contains novel classes absent from training, rendering class proportions and novel-class densities unidentifiable. To tackle this challenge, we propose the first identifiable semiparametric density ratio estimation framework. By introducing an overlap modeling mechanism between novel and known classes, our approach ensures identifiability without strong assumptions or prior knowledge, supported by rigorous theoretical guarantees. The method integrates maximum empirical likelihood estimation, asymptotically efficient confidence interval construction, a stable EM-based optimization algorithm, and a posterior-probability-based approximately optimal classifier. Extensive experiments on synthetic and real-world datasets demonstrate substantial improvements in both class proportion estimation accuracy and classification performance, consistently outperforming state-of-the-art methods across all benchmarks.

0 citationsRead paper

Recent publications

Latest Papers

LogicCBMs: Logic-Enhanced Concept-Based Learning

Dec 08, 2025

Existing concept bottleneck models (CBMs) rely on linear concept combinations, limiting expressivity and hindering modeling of complex semantic relationships. This paper proposes Logic-CBM, the first differentiable logic-driven CBM that embeds propositional logical operations (e.g., AND, OR, NOT) into an end-to-end trainable framework. Leveraging a neuro-symbolic design—integrating differentiable logic gates, gradient straight-through estimators, and joint optimization—it enables nonlinear logical reasoning over concepts. Evaluated on multiple benchmark and synthetic datasets, Logic-CBM achieves significant improvements in prediction accuracy, supports precise concept-level interventions, and preserves strong interpretability and model transparency. Its core contributions are threefold: (1) introducing the first differentiable logic-based concept learning paradigm; (2) overcoming the fundamental limitations of linear concept composition; and (3) unifying the structural rigor of symbolic logic with the learnability of deep neural networks.

0 citationsRead paper

Semiparametric Learning from Open-Set Label Shift Data

Sep 17, 2025

This paper addresses the open-set label shift problem, where the test distribution contains novel classes absent from training, rendering class proportions and novel-class densities unidentifiable. To tackle this challenge, we propose the first identifiable semiparametric density ratio estimation framework. By introducing an overlap modeling mechanism between novel and known classes, our approach ensures identifiability without strong assumptions or prior knowledge, supported by rigorous theoretical guarantees. The method integrates maximum empirical likelihood estimation, asymptotically efficient confidence interval construction, a stable EM-based optimization algorithm, and a posterior-probability-based approximately optimal classifier. Extensive experiments on synthetic and real-world datasets demonstrate substantial improvements in both class proportion estimation accuracy and classification performance, consistently outperforming state-of-the-art methods across all benchmarks.

0 citationsRead paper