Skip to content
View xxxxyu's full-sized avatar
🎯
Focusing
🎯
Focusing
44 followers · 26 following

Achievements

Achievement: StarstruckAchievement: Pull Sharkx2Achievement: Arctic Code Vault Contributor

Achievements

Achievement: StarstruckAchievement: Pull Sharkx2Achievement: Arctic Code Vault Contributor

Highlights

  • Pro

Organizations

@eesast @OpenBitSys @air-embodied-brain

Block or report xxxxyu

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
xxxxyu/README.md

Hi, I'm Xiangyu Li (李翔宇)

I'm a fifth-year Ph.D. candidate at the Institute for AI Industry Research (AIR), Tsinghua University, advised by Prof. Yunxin Liu and working closely with Prof. Ting Cao. I received my B.E. in Electronic Engineering from Tsinghua University in 2022.

I work on on-device AI and embodied AI, with a current focus on efficient deployment of embodied foundation models (VLAs and WAMs) and self-evolving physical intelligence. Homepage · RedNote

Selected work

  • Cosmos Lite [WIP] — quantized deployment of Cosmos 3 robot policies on a single 24 GB GPU.
  • ActProbe [WIP] — action-space probing for early failure detection of generative robot policies.
  • OxyGen — unified KV cache management for multi-task VLA inference.
  • Vec-LUT (MobiSys 2026 Best Paper Award Runner-Up) — parallel ultra-low-bit LLM inference on edge devices.
  • FlexNN (MobiCom 2024) — memory-adaptive DNN inference on edge devices.

Pinned Loading

  1. air-embodied-brain/Zetta-Embodiment air-embodied-brain/Zetta-Embodiment Public

    Zetta is an efficient closed-loop embodied harness for self-evolving physical intelligence. It evolves code-based runtime critics and recovery skills online while keeping the base policy frozen. Pr…

    Python 659 44

  2. air-embodied-brain/OxyGen air-embodied-brain/OxyGen Public

    Unified KV cache management for multi-task VLA inference.

    Python 11

  3. cosmos-lite cosmos-lite Public

    An efficient inference runtime for Cosmos 3 Nano / Edge robot policies

    Python 8

  4. OpenBitSys/vlut.cpp OpenBitSys/vlut.cpp Public

    [MobiSys 2026] On-device ultra-low-bit LLM inference with LUT.

    C++ 26 4

  5. FlexNN FlexNN Public

    [MobiCom 24] Adaptive DNN inference under memory constraints

    C 59 4

  6. xxxxyu.github.io xxxxyu.github.io Public

    Homepage built with zola

    Python