ZhiYuan(Aaron) Feng   |   冯志远

I am currently working on something incredibly exciting as a Ph.D. candidate in Computer Science at the Institute for Advanced Study, Tsinghua University, where I’m fortunate to be advised by Prof. Baining Guo. I received my Bachelor’s degree in Computer Science from the Qian Xuesen Honors College at Xi’an Jiaotong University.

My primary research interests lie in embodied AI, spatial reasoning, and multimodal large language models (MLLMs), with an emphasis on vision-language(-action) systems for robotic perception, understanding, and decision-making. Along the way, I’ve also explored related topics such as reinforcement learning (RL), large language models (LLMs), and retrieval-augmented generation (RAG). I’m also excited about bridging lab research with real-world scenarios to build systems that can genuinely support and improve everyday life. More broadly, I’m an AGI believer—and I’m genuinely excited to witness the arrival of ASI.

Email  /  Google Scholar  /  RedBook /  WeChat /  X

We survey how to turn abundant human videos into actionable supervision for VLA models, organizing methods into four categories and outlining key challenges for reliable real-world transfer.

We propose DLR, an information-theoretic multi-pattern RL framework that generates diverse, high-success manipulation trajectories for scalable VLA pretraining, improving transfer and showing better data-scaling than standard single-pattern RL.

We propose a diffusion policy optimization method that uses RL to autonomously generate smooth, low-variance long-horizon manipulation trajectories, enabling VLA training that outperforms models trained on human or Gaussian RL demonstrations on LIBERO.

We propose VITRA, a novel approach for pretraining Vision-Language-Action (VLA) models for robotic manipulation using large-scale, unscripted, real-world videos of human hand activities.

We propose MV-RoboBench, a benchmark for evaluating the multi-view spatial reasoning capabilities of VLMs in robotic manipulation.

HSSBench is a multilingual (UN six languages) benchmark of 13,000+ expert-curated samples designed to evaluate MLLMs’ cross-disciplinary, concept-to-vision reasoning in Humanities and Social Sciences, revealing clear gaps even in state-of-the-art models.

We introduce MIRA, a medical time-series foundation model pretrained on 454B+ time points that handles irregular sampling and missingness to deliver stronger zero-shot and fine-tuned forecasting across datasets and tasks.

We propose TransDiff, a diffusion-based single-view RGB-D depth completion method for accurate grasping of transparent objects.

We integrate evidential deep learning into DCRNN to provide sampling-free uncertainty quantification for spatiotemporal traffic forecasting, achieving improved predictive intervals measured by MIS.

This template is a modification to Jiayi Ni's website and Jon Barron's website.