Yuang Cao 曹宇昂

Cognitive science · Generative AI

Yuang Cao曹宇昂

Studying human cognition with AI as a research tool.

I am an undergraduate in Intelligent Medical Engineering at Southern University of Science and Technology (SUSTech). In Quanying Liu’s NCC Lab, I work on visual working memory, controllable visual stimuli and human evaluation of emotion descriptions.

IMAGE WHEEL CogSci 2025 · Fig. 1a
Image wheel from our paper: twelve generated object images arranged in a circle.
A small change in what we see. A measurable change in what we remember.
From our visual working memory study →

Research interests

Memory, visual stimuli, and emotion descriptions

Memory & perception

How does comparing what we remember with what we see change visual working memory?

Generative stimuli

How can we vary object features in a controlled way to ask more precise cognitive questions?

Human & model descriptions

Where do model-generated descriptions agree with human judgments, and where do they differ?

Selected publication

Visual working memory, beyond simple shapes

CogSci 2025 · Co-first author

Dimensions of Vulnerability in Visual Working Memory: An AI-Driven Approach to Perceptual Comparison

Yuang Cao*, Jiachen Zou*, Chen Wei†, Quanying Liu†

* Equal contribution · † Corresponding authors

We used generated object images to examine how perceptual comparisons distort visual working memory. The study compares image-based and dimension-based similarity, finding that similarity in object dimensions can also bias memory.

Study design & abstract
Figure 1 from the paper: image and dimension wheels, followed by the baseline and two perceptual comparison conditions.
Experimental design · Figure 1 from the paper

The paper tests memory under a baseline condition and two perceptual comparison conditions. Image wheels vary image similarity; dimension wheels vary object-dimension similarity. Both types of comparison can induce memory distortion, with visual dimensions more susceptible than semantic dimensions in this study.

Research notes

Emo-FilM: measurement items and subsequent studies

Interactive literature map · 2026

Emo-FilM: from 50 measurements to subsequent research

An interactive reading note that connects the dataset’s measurement items and fMRI to the data selections, analyses, and findings of subsequent studies.

Explore the interactive map ↗

Literature interpretation, with primary sources and explicit limits; not a new empirical study.

  1. Measurement

    Read all 50 items and their scale endpoints.

  2. Selection

    Highlight the data each study actually used.

  3. Interpretation

    Separate the findings from what they cannot establish.

Research experience

Visual memory and emotion descriptions

2024–2025

Visual working memory & perceptual comparison

CogSci 2025

Research in Quanying Liu’s NCC Lab, SUSTech

Research question
How do image similarity and object-dimension similarity affect memory distortion?
My contribution
Contributed to stimulus generation, behavioral study implementation, data analysis and manuscript preparation, in collaboration with the co-authors.
Output
Co-first-author paper at CogSci 2025.
2026–

Emotion descriptions in natural video

Exploratory research

Research in Quanying Liu’s NCC Lab, SUSTech

Research question
How should open-ended, model-generated emotion words be evaluated against human descriptions and judgments?
My contribution
Built clip preparation and subtitle-alignment workflows; worked on model-generated word sampling and online human-validation studies.
Current focus
Evaluating annotation quality and human agreement. This is ongoing work, without a published claim of a validated emotion space.

Education

Southern University of Science and Technology

B.Eng. student, Intelligent Medical Engineering · 2023–2027 (expected)

Explore the research map

A compact overview of the topics above. Select a topic to read more.

Get in touch

Contact

For questions about the paper or potential research collaborations, email is the best way to reach me.