Beauty is in the ELBO of the Beholder: A Variational Account of Processing Fluency in Face Perception
研究通过训练变分自编码器处理四个面部数据集，并评估其在芝加哥面部数据库上的表现，验证了面孔吸引力与感知流畅性之间的关系。
0 citationsRead paper
Selected work
Representative Papers
Re-assessing the evidence for mental rotation abilities in children using computational models
This study challenges the prevailing assumption that mental rotation (MR) emerges in infancy and questions the developmental timing of MR capacity in early childhood. Method: We constructed a developmental computational model grounded in embodied cognition to simulate behavioral performance of children aged 6 months to 5 years across three canonical MR tasks. The model systematically evaluated whether MR is a necessary cognitive mechanism underlying observed task behavior. Contribution/Results: We demonstrate that children’s performance is fully and precisely replicable using non-MR strategies—specifically, pixel-level stimulus matching—without invoking MR as a latent cognitive process. All empirical behavioral data are accurately fitted without assuming MR competence at any age. This work provides the first systematic computational evidence undermining the claim that MR is present in infancy, thereby challenging the empirical foundations of early-MR accounts and prompting a theoretical reassessment of mechanisms underlying spatial cognitive development in young children.
0 citationsRead paper
Combining complex Langevin dynamics with score-based and energy-based diffusion models
The sign problem—arising from complex-valued actions in quantum field theory and hindering Monte Carlo sampling via complex Langevin dynamics—remains a fundamental challenge. Method: This work introduces, for the first time, score-based and energy-based diffusion models into complex configuration space to implicitly learn and reconstruct the underlying non-equilibrium, non-real-valued probability distributions. Training data are generated via complex Langevin dynamics; both model classes are trained separately for distribution estimation and sample reconstruction. Theoretical analysis establishes their representational capacity for such complex, non-Hermitian distributions. Contribution/Results: Experiments demonstrate that the learned models accurately approximate the target complex Langevin distribution, achieving substantial improvements in sampling efficiency and numerical stability. This work establishes a novel machine-learning paradigm for tackling the sign problem in quantum field theory and extends the theoretical foundations and physical applicability of generative diffusion models to complex-probability systems.
0 citationsRead paper
Early Detection of Visual Impairments at Home Using a Smartphone Red-Eye Reflex Test
This study addresses the global challenge of limited access to specialized ophthalmologists and clinical infrastructure for early visual impairment screening in young children. We propose a novel, smartphone-based, home-administered Bruckner test leveraging red reflex imaging—marking the first mobile adaptation of this clinical method. Our approach comprises: (1) construction of a pediatric pupil image dataset annotated by board-certified ophthalmologists; and (2) development and deployment of a lightweight deep neural network model enabling real-time image acquisition and automated abnormality classification. Key contributions include: complete elimination of dedicated hardware, enabling contactless, low-barrier, at-home screening; 90% classification accuracy on an independent test set; and empirical identification of optimal imaging conditions (e.g., ambient illumination, camera-to-subject distance, and angle). This framework significantly enhances accessibility and scalability of early childhood vision screening worldwide.
0 citationsRead paper
From Kicking to Causality: Simulating Infant Agency Detection with a Robust Intrinsic Reward
Standard reinforcement learning relies on correlational rewards, rendering it brittle in ecologically realistic, noisy environments where agents struggle to robustly identify the causal effects of their own actions—yet human infants develop such causal agency early in development. To address this, we propose a causally grounded intrinsic reward mechanism that defines the Causal Action Influence Score (CAIS), quantifying the causal effect of an agent’s actions on perceptual outcomes via the 1-Wasserstein distance, and integrates prediction-error-driven surprise signals for noise filtering. This work is the first to systematically incorporate causal inference into intrinsic reward design, successfully reproducing the psychological phenomenon of “extinction burst.” Experiments demonstrate that, under strong external interference where conventional correlational rewards fail completely, our method reliably identifies the agent’s causal efficacy, significantly enhancing robustness and policy acquisition in ecologically valid settings.
0 citationsRead paper
Recent publications
Latest Papers
Beauty is in the ELBO of the Beholder: A Variational Account of Processing Fluency in Face Perception
研究通过训练变分自编码器处理四个面部数据集，并评估其在芝加哥面部数据库上的表现，验证了面孔吸引力与感知流畅性之间的关系。
0 citationsRead paper
Re-assessing the evidence for mental rotation abilities in children using computational models
This study challenges the prevailing assumption that mental rotation (MR) emerges in infancy and questions the developmental timing of MR capacity in early childhood. Method: We constructed a developmental computational model grounded in embodied cognition to simulate behavioral performance of children aged 6 months to 5 years across three canonical MR tasks. The model systematically evaluated whether MR is a necessary cognitive mechanism underlying observed task behavior. Contribution/Results: We demonstrate that children’s performance is fully and precisely replicable using non-MR strategies—specifically, pixel-level stimulus matching—without invoking MR as a latent cognitive process. All empirical behavioral data are accurately fitted without assuming MR competence at any age. This work provides the first systematic computational evidence undermining the claim that MR is present in infancy, thereby challenging the empirical foundations of early-MR accounts and prompting a theoretical reassessment of mechanisms underlying spatial cognitive development in young children.
0 citationsRead paper
Combining complex Langevin dynamics with score-based and energy-based diffusion models
The sign problem—arising from complex-valued actions in quantum field theory and hindering Monte Carlo sampling via complex Langevin dynamics—remains a fundamental challenge. Method: This work introduces, for the first time, score-based and energy-based diffusion models into complex configuration space to implicitly learn and reconstruct the underlying non-equilibrium, non-real-valued probability distributions. Training data are generated via complex Langevin dynamics; both model classes are trained separately for distribution estimation and sample reconstruction. Theoretical analysis establishes their representational capacity for such complex, non-Hermitian distributions. Contribution/Results: Experiments demonstrate that the learned models accurately approximate the target complex Langevin distribution, achieving substantial improvements in sampling efficiency and numerical stability. This work establishes a novel machine-learning paradigm for tackling the sign problem in quantum field theory and extends the theoretical foundations and physical applicability of generative diffusion models to complex-probability systems.
0 citationsRead paper
Early Detection of Visual Impairments at Home Using a Smartphone Red-Eye Reflex Test
This study addresses the global challenge of limited access to specialized ophthalmologists and clinical infrastructure for early visual impairment screening in young children. We propose a novel, smartphone-based, home-administered Bruckner test leveraging red reflex imaging—marking the first mobile adaptation of this clinical method. Our approach comprises: (1) construction of a pediatric pupil image dataset annotated by board-certified ophthalmologists; and (2) development and deployment of a lightweight deep neural network model enabling real-time image acquisition and automated abnormality classification. Key contributions include: complete elimination of dedicated hardware, enabling contactless, low-barrier, at-home screening; 90% classification accuracy on an independent test set; and empirical identification of optimal imaging conditions (e.g., ambient illumination, camera-to-subject distance, and angle). This framework significantly enhances accessibility and scalability of early childhood vision screening worldwide.
0 citationsRead paper
From Kicking to Causality: Simulating Infant Agency Detection with a Robust Intrinsic Reward
Standard reinforcement learning relies on correlational rewards, rendering it brittle in ecologically realistic, noisy environments where agents struggle to robustly identify the causal effects of their own actions—yet human infants develop such causal agency early in development. To address this, we propose a causally grounded intrinsic reward mechanism that defines the Causal Action Influence Score (CAIS), quantifying the causal effect of an agent’s actions on perceptual outcomes via the 1-Wasserstein distance, and integrates prediction-error-driven surprise signals for noise filtering. This work is the first to systematically incorporate causal inference into intrinsic reward design, successfully reproducing the psychological phenomenon of “extinction burst.” Experiments demonstrate that, under strong external interference where conventional correlational rewards fail completely, our method reliably identifies the agent’s causal efficacy, significantly enhancing robustness and policy acquisition in ecologically valid settings.
0 citationsRead paper