Ciao! I'm Pengsheng Guo.

I’m a Tech Lead Manager at Apple, working at the intersection of cutting-edge machine learning research and large-scale product deployment. I led the development and launch of the 2-bit on-device Apple Foundation Model, and contributed key features to Apple Vision Pro, including scene understanding and Persona.

I conduct research in multiple domains, including multimodal LLM, diffusion model (image, video, 3D), neural rendering, multi-task learning, neural architecture search, and network optimization.

I work closely with Professor Alex Schwing on generative models. Previously, I worked with Professor David Held on Reinforcement Learning in Robotics during my graduate studies at CMU Robotics Institute. I worked with Professor I-Ming Chen on Robotic Manipulation during undergrad at NTU, Singapore.

Tech Report

Apple's On-Device and Server Foundation Language Models
_{Apple Machine Learning Research Blog, 2025}

Publications

(* indicates equal contribution)

Variational Rectified Flow Matching
_{Pengsheng Guo, Alex Schwing}
_{International Conference on Machine Learning (ICML), 2025}

CommVQ: Commutative Vector Quantization for KV Cache Compression
_{Junyan Li, Yang Zhang, Muhammad Yusuf Hassan, Talha Chafekar, Tianle Cai, Zhile Ren, Pengsheng Guo, Foroozan Karimzadeh, Colorado Reed, Chong Wang, Chuang Gan}
_{International Conference on Machine Learning (ICML), 2025}

On Inductive Biases That Enable Generalization in Diffusion Transformers
_{Jie An, De Wang, Pengsheng Guo, Jiebo Luo, Alex Schwing}
_{arXiv, 2024}

StableDreamer: Taming Noisy Score Distillation Sampling for Text-to-3D
_{Pengsheng Guo, Hans Hao, Adam Caccavale, Zhongzheng Ren, Edward Zhang, Qi Shan, Aditya Sankar, Alexander G. Schwing, Alex Colburn, Fangchang Ma}
_{arXiv, 2023}

CVRecon: Rethinking 3d geometric feature learning for neural reconstruction
_{Ziyue Feng, Liang Yang, Pengsheng Guo, Bing Li}
_{Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV), 2023}

GAUDI: A Neural Architect for Immersive 3D Scene Generation
_{Miguel Angel Bautista*, Pengsheng Guo*, Samira Abnar, Walter Talbott, Alexander Toshev, Zhuoyuan Chen, Laurent Dinh, Shuangfei Zhai, Hanlin Goh, Daniel Ulbricht, Afshin Dehghan, Josh Susskind}
_{Advances in Neural Information Processing Systems (NeurIPS), 2022}

Fast and Explicit Neural View Synthesis
_{Pengsheng Guo, Miguel Angel Bautista, Alex Colburn, Liang Yang, Daniel Ulbricht, Joshua M. Susskind, Qi Shan}
_{Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision (WACV), 2022}

MetricOpt: Learning to Optimize Black-Box Evaluation Metrics
_{Chen Huang, Shuangfei Zhai, Pengsheng Guo, Josh Susskind}
_{Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2021}

Learning to Branch for Multi-Task Learning
_{Pengsheng Guo, Chen-Yu Lee, Daniel Ulbricht}
_{International Conference on Machine Learning (ICML), 2020}

Adaptive variance for changing sparse-reward environments
_{Xingyu Lin, Pengsheng Guo, Carlos Florensa, David Held}
_{International Conference on Robotics and Automation (ICRA), 2019}

Deploying social robots as teaching aid in pre-school K2 classes: A proof-of-concept study
_{Albert Causo; Phyo Zin Win; Pengsheng Guo; I.-Ming Chen}
_{International Conference on Robotics and Automation (ICRA), 2017}