Luong Thai Khang

B.Sc. in Computer Science · Graduated July 2026

Bandits, Reinforcement Learning, and Decision-making under Uncertainty.

I graduated with a B.Sc. in Computer Science from Hanoi University of Science and Technology in July 2026. My research focuses on bandits, reinforcement learning, offline/robust decision-making, and statistical learning theory. I am first author of papers at ICML 2026 and NeurIPS 2026 (Spotlight), and am seeking Master’s/Ph.D. opportunities and research collaborations.

Download CV PDF Email GitHub

Research interests

Bandit Theory Reinforcement Learning Pure Exploration Best-Arm Identification 1-Bit Mean Estimation Offline Contextual Bandits Distributionally Robust Optimization Risk-Aware Decision Making

Publications and manuscripts

  1. Variance Driven Exploration: A Provable and Efficient Methodology for Pure Exploration in Highly Stochastic Environments
    Khang Luong, Nam Nguyen, Hoang Ta, Hung The Tran, Tuan Quang Dam.
    ICML 2026 (Regular). OpenReview · arXiv
  2. Nearly Optimal Fixed-Confidence Best-Arm Identification with 1-Bit Feedback
    Khang Luong, Dinh Thai Son, Hoang Ta, Hung The Tran, Tuan Quang Dam.
    NeurIPS 2026 (Spotlight). OpenReview · arXiv
  3. Minimax Optimality for Factored Optimal-Transport Distributionally-Robust Policy Learning in Contextual Bandits
    Khang Luong, Trung Nguyen, Hoang Ta, Tuan Quang Dam, Hung Tran-The.
    ICLR 2027 Conference Submission.
  4. What Action Geometry Buys: A Design Space for Optimal-Transport Bellman Backups
    Trung Nguyen, Khang Luong, Tuan Quang Dam.
    ICLR 2027 Conference Submission.
  5. Instance-dependent β-Optimality for Best-Arm Identification under One-Bit Threshold Feedback
    Dinh Thai Son, Khang Luong, Hoang Ta, Tuan Quang Dam.
    AISTATS 2027 Conference Submission.

Education

Hanoi University of Science and Technology

B.Sc. in Computer Science · GPA: 3.73/4.0

2022 – Jul. 2026

Language

English: IELTS 7.0

Experience

VinDynamics

AI Intern, Robotics. Worked on perception and control pipelines for robotic grasping, hand–object interaction, and vision-based teleoperation.

Jun. 2025 – May 2026

FPT Quantum AI & Cyber Security Institute

Research Member. Conducting research on offline policy evaluation and learning for contextual bandits.

Jun. 2026 – Present

Selected projects

Variance Driven Exploration (VarDE)

Hanoi University of Science and Technology (HUST)

  • Derived influence-weighted variance-driven sampling rules for exploration frameworks, providing theoretical guarantees on variance decay and simple regret.
  • Applied VarDE to Best-Arm Identification, Monte Carlo Tree Search, and Best-Policy Identification.

Best-Arm Identification with 1-Bit Feedback

Hanoi University of Science and Technology (HUST)

  • Developed a BAI algorithm with 1-bit mean estimation, matching the worst-case lower bound for 1-bit feedback.
  • Currently extending the analysis to instance-dependent bounds and asymptotically optimal parametric algorithms.

Offline Policy Evaluation and Learning for Contextual Bandits

FPT Quantum AI & Cyber Security Institute

  • Study distributionally robust offline contextual bandits under context shift and reward shift.
  • Develop pessimistic offline policy evaluation and learning methods based on optimal-transport DRO objectives.
  • Analyze finite-sample guarantees under limited coverage and offline-data constraints.

Vision-Based Robotic Hand Teleoperation and Autonomous Manipulation

VinDynamics

  • Developed a vision-based teleoperation system for dexterous robotic hand control.
  • Worked on perception and control pipelines for robotic grasping and hand–object interaction.
  • Integrated visual perception with control policies for autonomous object manipulation.

Contact

I am interested in conversations with faculty, postdocs, Ph.D. students, and collaborators working on machine learning theory, bandits, reinforcement learning, offline/robust decision-making, and statistical learning theory.

luongkhang448@gmail.com github.com/luongkhang04 Download CV PDF