Pengcheng Pan

I am a Project Researcher at the International Research Center for Neurointelligence (IRCN), The University of Tokyo. I received my Ph.D. in Information Science and Technology from The University of Tokyo in 2026.

My research focuses on human-centered active vision, hard attention, and brain-inspired AI. I study vision models that actively acquire visual evidence through sequential gaze-like observations, rather than processing the entire image passively in a single step. My broader goal is to build AI systems whose perception, attention, and decision processes are more interpretable, controllable, and informed by human perceptual behavior.

Research Interests

• Active vision and hard attention
• Human-like scanpaths and gaze behavior
• Brain-inspired and cognitive AI models
• Interpretable and privacy-aware visual recognition
• Multimodal AI and vision-language systems
• Predictive coding and cognitive feelings
• Human-AI interaction and human-centered AI

Research Profile

My doctoral research investigated hard-attention vision models inspired by human eye movements. These models observe local image regions sequentially, update internal recurrent states, and decide where to look next. I evaluate such models not only by classification accuracy, but also by their scanpath behavior, fixation patterns, spatial exploration, and similarity to human gaze data.

This line of work connects computer vision, reinforcement learning, cognitive science, and neuroscience. It also provides a practical route toward interpretable and privacy-aware AI systems, where the visual evidence used by a model can be inspected, audited, and intervened on during inference.

Current Position

Project Researcher, International Research Center for Neurointelligence (IRCN), The University of Tokyo
May 2026 - Present

I work on AI models inspired by cognitive science and neuroscience, with a focus on visual attention, active perception, predictive coding, cognitive feelings, and multimodal learning.

Experience

Project Researcher IRCN, The University of Tokyo

Research on brain-inspired AI, active vision, predictive coding, cognitive feelings, and multimodal learning.

2026-Present

Machine Learning Engineer Intern Corpy & Co., Inc.

Implemented and evaluated machine learning models, analyzed experimental results, and built literature information extraction pipelines using the Semantic Scholar API and OpenAI API.

2023-2024

Social Robot Research Intern Tata Consultancy Services Japan

Worked on multimodal human-robot interaction using facial expression recognition, robot vision, dialogue generation, and generative AI tools.

2022

Research Intern Chulalongkorn University

Studied model interpretability for audio recognition using YAMNet and Grad-CAM visualization.

Feb-Mar 2023

Education

Ph.D. in Information Science and Technology The University of Tokyo

Department of Mechano-Informatics, Graduate School of Information Science and Technology. Dissertation: Human-Inspired, Affect Sensitive Hard Attention Model Yields Human-Like Gaze in Image Classification.

2023-2026

M.S. in Information Science and Technology The University of Tokyo

Department of Mechano-Informatics, Graduate School of Information Science and Technology. Thesis: Abstraction and conceptualization of spatial information through task-specific active perception.

2021-2023

Graduate Student Tohoku University

Department of Quantum Science and Energy Engineering, School of Engineering.

2019-2021

Exchange Student / Research Student Tohoku University

School of Engineering, Department of Mechanical and Aerospace Engineering.

2018-2019

B.Eng. in Mechanical and Electronic Engineering Jiangnan University

2015-2019

Selected Projects

• EVA: Human-Aligned Hard-Attention Vision
• MRAM: Multi-Level Recurrent Attention Model
• Peripheral Sweet Spot for Human-like Scanpaths
• Human-in-the-Loop Hard Attention for Privacy-Aware Active Vision

Selected Publications

• Pan Pengcheng, Shogo Yonekura, Yasuo Kuniyoshi. “Debiasing Central Fixation Confounds Reveals a Peripheral ‘Sweet Spot’ for Human-like Scanpaths in Hard-Attention Vision.” CogSci 2026.
• Pan Pengcheng, Shogo Yonekura, Yasuo Kuniyoshi. “Emergence of Fixational and Saccadic Movements in a Multi-Level Recurrent Attention Model for Vision.” ICONIP 2025.
• Pan Pengcheng, Shogo Yonekura, Yasuo Kuniyoshi. “Bayesian Glimpse Control Boosts Active Vision RL on Atari Pong.” IWAI 2025, Poster & Spotlight.
• Pan Pengcheng, Shogo Yonekura, Yasuo Kuniyoshi. “Connecting Dense and Sparse Networks in Recurrent Independent Mechanisms.” RSJ 2022, Oral.
• Pengcheng Pan, Shogo Yonekura, Yasuo Kuniyoshi. “Human-in-the-Loop Hard Attention for Privacy-Aware Active Vision.” JSAI 2026.