Selected work

Representative Papers

Sensorimotor features of self-awareness in multimodal large language models

May 25, 2025

This study investigates whether multimodal large language models (MLLMs) can spontaneously develop bodily self-awareness solely through embodied sensorimotor interaction. We embed an MLLM in an autonomous mobile robot that explores its environment and learns closed-loop behaviors exclusively from real-time multimodal sensory inputs—vision, touch, proprioception, and vestibular signals—without any explicit supervision or pre-defined self-models. We systematically evaluate the model’s capabilities in environmental recognition, self-discrimination, and motor prediction. Our key contributions are threefold: (1) First empirical evidence that MLLMs hierarchically emergent bodily self-awareness in a fully unsupervised, embodied setting; (2) Causal insights—derived via structural equation modeling and sensory ablation experiments—into how multisensory integration, temporal memory, and hierarchical internal representations jointly enable self-awareness; and (3) Demonstration that structured and episodic memory are essential for coherent self-referential reasoning, along with identification of critical sensory modalities and their functional redundancy relationships.

0 citationsRead paper

Detection of Endangered Deer Species Using UAV Imagery: A Comparative Study between Efficient Deep Learning Approaches

May 14, 2025International Conference on Unmanned Aircraft Systems

Detecting endangered deer species (e.g., marsh deer) in drone-captured imagery remains challenging due to their small object size, low spatial占比, and severe occlusion by dense vegetation, leading to degraded detection performance. Method: This paper proposes a YOLO-based framework enhanced with instance segmentation, integrating a lightweight segmentation head into YOLOv11 and RT-DETR variants. We construct a high-fidelity, pixel-level mask-annotated dataset specifically for wetland deer detection and design a wetland-adapted drone image augmentation pipeline. Contribution/Results: Experimental results demonstrate a 12.3% improvement in mean Average Precision (mAP) under heavy occlusion and complex backgrounds. The method significantly enhances localization robustness and classification accuracy for small targets, offering an efficient, scalable, and deployable solution for automated monitoring of endangered cervids in natural habitats.

0 citationsRead paper

Recent publications

Latest Papers

Sensorimotor features of self-awareness in multimodal large language models

May 25, 2025

This study investigates whether multimodal large language models (MLLMs) can spontaneously develop bodily self-awareness solely through embodied sensorimotor interaction. We embed an MLLM in an autonomous mobile robot that explores its environment and learns closed-loop behaviors exclusively from real-time multimodal sensory inputs—vision, touch, proprioception, and vestibular signals—without any explicit supervision or pre-defined self-models. We systematically evaluate the model’s capabilities in environmental recognition, self-discrimination, and motor prediction. Our key contributions are threefold: (1) First empirical evidence that MLLMs hierarchically emergent bodily self-awareness in a fully unsupervised, embodied setting; (2) Causal insights—derived via structural equation modeling and sensory ablation experiments—into how multisensory integration, temporal memory, and hierarchical internal representations jointly enable self-awareness; and (3) Demonstration that structured and episodic memory are essential for coherent self-referential reasoning, along with identification of critical sensory modalities and their functional redundancy relationships.

0 citationsRead paper

Detection of Endangered Deer Species Using UAV Imagery: A Comparative Study between Efficient Deep Learning Approaches

May 14, 2025International Conference on Unmanned Aircraft Systems

Detecting endangered deer species (e.g., marsh deer) in drone-captured imagery remains challenging due to their small object size, low spatial占比, and severe occlusion by dense vegetation, leading to degraded detection performance. Method: This paper proposes a YOLO-based framework enhanced with instance segmentation, integrating a lightweight segmentation head into YOLOv11 and RT-DETR variants. We construct a high-fidelity, pixel-level mask-annotated dataset specifically for wetland deer detection and design a wetland-adapted drone image augmentation pipeline. Contribution/Results: Experimental results demonstrate a 12.3% improvement in mean Average Precision (mAP) under heavy occlusion and complex backgrounds. The method significantly enhances localization robustness and classification accuracy for small targets, offering an efficient, scalable, and deployable solution for automated monitoring of endangered cervids in natural habitats.

0 citationsRead paper