Enguang Wang

Enguang Wang

I am currently a third-year Master’s student at the College of Computer Science, Nankai University, advised by Prof. Xialei Liu and Prof. Ming-Ming Cheng. Prior to this, I received my bachelor’s degree from the School of Artificial Intelligence, Xidian University.

My research interests lie in Multi-modal Large Language Models (MLLMs) and Representation Learning. I firmly believe that representation is the cornerstone of AI systems, and I am dedicated to exploring robust and efficient representation learning approaches.

I am currently an Applied Research Intern at ByteDance, where I focus on the application and system design of Agentic MLLMs, specifically exploring memory mechanisms and Agentic RL.

My current research remains dedicated to enhancing the efficiency, safety, and foundational representation of MLLMs:

  • Visual Latent Reasoning: I am exploring collaborative optimization of state-conditioned and reasoning latents to achieve implicit “Chain-of-Thought” reasoning. By enhancing latent information density through decoder-based refinement, this approach aims to boost reasoning performance while significantly reducing inference costs. My experiments also investigate latent degradation issues in recent latent reasoning methods.
  • Efficient Capability Alignment: I am investigating methods to rectify and activate model capabilities using a minimal set of anchor tokens, aiming to bridge the performance gap between base models and RL-aligned models. This research focuses on “switching” models into high-reasoning states without the prohibitive cost of full-scale Reinforcement Learning.
  • Representation Learning: Designing Text-Conditioned Encoders to improve cross-modal alignment and foundational visual representation for MLLMs.
  • AI Safety: Investigating adversarial attacks and robustness for agents.

I am expected to graduate in 2026.06. I am currently actively seeking research-oriented job opportunities (Research Scientist/Engineer). If you have potential openings or are seeking any form of academic cooperation, please feel free to email me.