cv · computational cognitive modeling

Computational cognitive modeling CV — emphasizes POMDP / RL / Bayesian modeling of human decision-making, social cognition, and neural decoding. Targets: DeepMind, Anthropic, Meta AI, computational-neuroscience labs.

Qijun (Bell) Zhong

Undergraduate · Sun Yat-sen University · Embodied AI & Computational Cognitive Science

zhongqj7@mail2.sysu.edu.cn · zhongqijunbell@gmail.com · gaigebell.github.io · github.com/gaigebell · LinkedIn · Google Scholar

Guangzhou, China

Education

Sun Yat-sen University 2024 – 2028
B.Eng. in Computer Science and Technology (expected) · School of Computer Science and Engineering
Sun Yat-sen University — Department of Psychology 2023 – 2024
(transferred to Computer Science) · Department of Psychology

Computational Modeling Projects

Planning under Goal Uncertainty 2024-08 – 2026-04
Conceptualization, modeling, result analysis
  • Proposed two-phase MDP to model optimal strategy for decision-making under goal uncertainty, separating goal inference from goal-conditioned planning.
  • Model comparison(using BIC) across optimal-policy, feature-based heuristic (linear regression), and mixed-strategy (logistic regression) models; identified which class of model best captures residual behavioral variance.
  • Vectorized single-goal heuristic predictors to scale the model to the multi-goal setting without retraining.
  • Parameter fitting against human behavior data; the framework generalizes across task families.
Choosing Cooperation Partners: Learning from Social Cognition 2024-04 – 2024-08
Modeling, result analysis
  • Built a reinforcement-learning framework for partner-selection under the competence-vs-warmth trade-off in social cognition.
  • Proposed a linear symbolic model to capture human's prior preference for warmth or competence.
  • Parallelized parameter search for fitting candidate computational models; precision 0.1 → 0.001.
  • Modeled how humans balance instrumental (competence) and affective (warmth) cues when choosing cooperation partners.
Reconstructing Chinese Semantic Representation from Brain Recordings 2024-04 – 2026-08
Project Leader
  • Neural decoding pipeline — reconstructing Chinese-language semantic representations from non-invasive EEG/fMRI signals using deep learning.
  • End-to-end stack — preprocessing, training, evaluation, inference.
  • Adopted CNN, RNN, and Transformer architectures to capture spatial and temporal patterns in neural data.
  • Adopted information bottlenecks to allign neural representations with semantic features extracted from LLM.
  • Integrated beam search decoding into LLM text generation to improve inference quality.
  • DDP-based multi-node multi-GPU training; compressed training time from 4 days → 1.5 days on identical model & data.
  • Bridges brain-computer interface (BCI) and computational cognitive modeling of language.

Publications

Manuscripts in preparation — see /projects/ for current work.

Skills

Programming Languages: Python (PyTorch, NumPy, scikit-learn), C/C++, Matlab
Tools & Platforms: Linux, Git, Docker, LaTeX, ROS, WebSocket, Notion, Obsidian
Domain: Reinforcement Learning, Deep Learning, Robot Manipulation (Franka), EEG/fMRI signal processing, Bayesian modeling
Languages: Mandarin (native), English (CET-6 605/ IELTS in progress)