Qing Li

Research scientist at #MultimodalLearning, #LLMAgents, #NeSy. Prev. Ph.D. @UCLA, BS @USTC, Intern @GoogleAI, @MSFTResearch, @AmazonScience.

[Qing Li @Sealiqing]

Exactly what I am doing right now!😎 Check my personal knowledge base for both me and my research agent: https://t.co/zIjZPSHj7U, built with Obsidian and Claude Code.

[Andrej Karpathy @karpathy]

LLM Knowledge Bases

Something I'm finding very useful recently: using LLMs to build personal knowledge bases for various topics of research interest. In this way, a large fraction of my recent token throughput is going less into manipulating code, and more into manipulating knowledge (stored as markdown and images). The latest LLMs are quite good at it. So:

Data ingest:
I index source documents (articles, papers, repos, datasets, images, etc.) into a raw/ directory, then I use an LLM to incrementally "compile" a wiki, which is just a collection of .md files in a directory structure. The wiki includes summaries of all the data in raw/, backlinks, and then it categorizes data into concepts, writes articles for them, and links them all. To convert web articles into .md files I like to use the Obsidian Web Clipper extension, and then I also use a hotkey to download all the related images to local so that my LLM can easily reference them.

IDE:
I use Obsidian as the IDE "frontend" where I can view the raw data, the the compiled wiki, and the derived visualizations. Important to note that the LLM writes and maintains all of the data of the wiki, I rarely touch it directly. I've played with a few Obsidian plugins to render and view data in other ways (e.g. Marp for slides).

Q&A:
Where things get interesting is that once your wiki is big enough (e.g. mine on some recent research is ~100 articles and ~400K words), you can ask your LLM agent all kinds of complex questions against the wiki, and it will go off, research the answers, etc. I thought I had to reach for fancy RAG, but the LLM has been pretty good about auto-maintaining index files and brief summaries of all the documents and it reads all the important related data fairly easily at this ~small scale.

Output:
Instead of getting answers in text/terminal, I like to have it render markdown files for me, or slide shows (Marp format), or matplotlib images, all of which I then view again in Obsidian. You can imagine many other visual output formats depending on the query. Often, I end up "filing" the outputs back into the wiki to enhance it for further queries. So my own explorations and queries always "add up" in the knowledge base.

Linting:
I've run some LLM "health checks" over the wiki to e.g. find inconsistent data, impute missing data (with web searchers), find interesting connections for new article candidates, etc., to incrementally clean up the wiki and enhance its overall data integrity. The LLMs are quite good at suggesting further questions to ask and look into.

Extra tools:
I find myself developing additional tools to process the data, e.g. I vibe coded a small and naive search engine over the wiki, which I both use directly (in a web ui), but more often I want to hand it off to an LLM via CLI as a tool for larger queries.

Further explorations:
As the repo grows, the natural desire is to also think about synthetic data generation + finetuning to have your LLM "know" the data in its weights instead of just context windows.

TLDR: raw data from a given number of sources is collected, then compiled by an LLM into a .md wiki, then operated on by various CLIs by the LLM to do Q&A and to incrementally enhance the wiki, and all of it viewable in Obsidian. You rarely ever write or edit the wiki manually, it's the domain of the LLM. I think there is room here for an incredible new product instead of a hacky collection of scripts.

[Qing Li @Sealiqing]

Cool work! When will a humanoid robot win a tennis Grand Slam?🤔

[Zhikai Zhang @Zhikai273]

🎾Introducing LATENT: Learning Athletic Humanoid Tennis Skills from Imperfect Human Motion Data

Dynamic movements, agile whole-body coordination, and rapid reactions. A step toward athletic humanoid sports skills.

Project: https://t.co/MFy2NIOsrn
Code: https://t.co/A7B5H8PIBh

[Zhikai Zhang @Zhikai273]

🎾Introducing LATENT: Learning Athletic Humanoid Tennis Skills from Imperfect Human Motion Data

Dynamic movements, agile whole-body coordination, and rapid reactions. A step toward athletic humanoid sports skills.

Project: https://t.co/MFy2NIOsrn
Code: https://t.co/A7B5H8PIBh

[Siyuan Huang @siyuanhuang95]

You might have seen the WuBOT performing at the 2026 Spring Festival Gala; however, most high-dynamic extreme motions you see are executed by overfitted tracking policies. Until now, training a unified policy capable of performing various extreme motions with a high success rate remained an unsolved challenge.

We spent an entire year digging into the barrier between general tracking and extreme physical behaviors. After burning through dozens of G1 robots, we finally identified the bottleneck of learning and physical executability.

With these discoveries, we developed OmniXtreme: the first general policy that can execute diverse extreme motions, including consecutive flips, extreme balancing, and even breakdancing with rapid contact switches!

This capability is achieved by pre-training a flow-based generative control policy and then post-training with actuation-aware residual RL for complex physical dynamics—a step we found critical for successful real-world transfer.

This work is a joint collaboration with @UnitreeRobotics. Together, we are pushing the physical limits of humanoid robots. It is incredibly exciting to see a general "robot gymnast" and "robot breakdancer" come to life! It was also our first time publishing a paper with XingXing, which was an enlightening experience.

The model checkpoints are now released—we welcome you to play with them! 📦

📄 Paper: https://t.co/ySKLac8w6i
🌐 Project: https://t.co/ortlLW24JB
💻 Code: https://t.co/IJ4XHYT4Qj

[Siyuan Huang @siyuanhuang95]

🥰Super excited that SceneWeaver (https://t.co/uAutDQ6Ay5) won the best paper award at the IROS25 RoboGen workshop. SceneWeaver provides an agentic framework for tool-based 3D scene generation, given a language description as input, you can generate or edit a corresponding details with lots of details.

[Siyuan Huang @siyuanhuang95]

🎉🎉 UniFP won the best paper award at CoRL25, we truly appreciate that the award committee and organizers really care about the fundamental control problem that can be solved by learning-based approach!

Meanwhile, I felt fully engaged during the entire conference, from keynote to oral to poster, which is a truly pleasure experience after many years research. The organization and vibe of #CoRL2025 are wonderful!

Congratulations to all the collaborators, the code will be fully open-sourced very soon!

[Qing Li @Sealiqing]

@TsingYoga Nice!!! It will be very useful for agentic RL research.

[Yujia Qin @TsingYoga]

The tool/env infra behind UI-TARS-2 is open-sourced. Enjoy the All-in-One Agent Sandbox!🥳

https://t.co/Lzl2dPqpQ7
https://t.co/p4LIFMp9Hr

[Qing Li @Sealiqing]

Nice!!! It will be very useful for agentic RL research.

[Yujia Qin @TsingYoga]

The tool/env infra behind UI-TARS-2 is open-sourced. Enjoy the All-in-One Agent Sandbox!🥳

https://t.co/Lzl2dPqpQ7
https://t.co/p4LIFMp9Hr

[Qing Li @Sealiqing]

@liuziwei7 @lmmslab Nice work!

[Ziwei Liu @liuziwei7]

🔥LLaVA-OneVision upgraded to V1.5🔥

We @lmmslab present 🌋LLaVA-OV-1.5🌋, a fully open framework for democratized multimodal training

* Superior Performance surpassing Qwen2.5-VL
* High-Quality Data at Scale
* Ultra-Efficient Training Framework

- Repo: https://t.co/1Mm6Mq5jqR

[Yukang Chen @yukangchen_]

Video understanding isn't just recognizing —it demands reasoning across thousands of frames.

Meet Long-RL🚀 Highlights:
🧠 Dataset: LongVideo-Reason — 52K QAs with reasoning.
⚡ System: MR-SP - 2.1× faster RL for long videos.
📈 Scalability: Hour-long videos (3,600 frames) RL on a single node (8×A100s).

🖼️📝🎵 RL training for video, text, audio — works with VILA, Qwen series, and image/video generation models 🎨🎬

📄 Paper: https://t.co/vbU5n0w0go
🎥 Demo: https://t.co/3wCv5TJsTa
💻 Code: https://t.co/K9U4fl3HHc

[Qing Li @Sealiqing]

@ChuanxiaZ @Oxford_VGG @NTUsg Big congrats!

[Hangxin Liu @HangxinLiu]

5 km run, zero mistakes. 🏃‍♂️🤖 Warming up for the Beijing Robot Half Marathon. #RobotMarathon

[Siyuan Huang @siyuanhuang95]

🤖🤖🤖 Following RoboVerse, we introduce another work focused on Robotic Tactile Simulation - Taccel Simulator. Taccel is a high-performance simulation platform for vision-based tactile sensors and robots.

🚀🚀🚀 Boosted by Nvidia Warp, we optimize Taccel with highly parallelized simulations and support 900fps simulation with 4k+ parallel training envs.

🤝🤝🤝 Taccel is designed with user-friendly APIs and is easy to use. We open-sourced all the code and documentation. Feel free to try!

Project: https://t.co/AT0G7MGzqX
Preprint: https://t.co/wSMUqBCwQB
Code: https://t.co/H5CxVjg5Ke

[ICLR @iclr_conf]

Announcing the keynote speakers for #ICLR2025! Speakers will cover topics ranging from foundational advances in language models, AI safety, open-ended learning, and the nature of intelligence itself. https://t.co/nL73E3KtbG

[Siyuan Huang @siyuanhuang95]

RoboVerse features tremendous solid progress in robotics. Including
- MetaSim, a configuration system and a universal interface to align current robotic simulators
- RoboVerse Dataset , a large-scale, high-quality synthetic dataset that integrates most of the current synthetic robotics dataset
-RoboVerse Benchmark, a standardized benchmark across simulators that can test most SOTA models fairly
- Hybrid Simulation, you can compose the physics simulator and renderer freely, achieving flexible and high-fidelity simulation

All code, datasets, and wiki are now fully open-sourced. Welcome to try and play through https://t.co/5wIc78mfxe

[Yixin Chen @_yixinchen]

📢📢📢Excited to announce the 5th Workshop on 3D Scene Understanding for Vision, Graphics, and Robotics at #CVPR2025! Expect our awesome speakers and challenges on multi-modal 3D scene understanding and reasoning. 🎉🎉🎉@CVPR

Learn more at https://t.co/hZEQr5qKxu.

[Yixin Chen @_yixinchen]

🚀 How to reconstruct 3D scenes with decomposed objects from sparse inputs?

Check out *DPRecon* (https://t.co/Sfp1yPoqxC) at #CVPR2025—it recovers all objects, achieves photorealistic mesh rendering, and supports text-based geometry & appearance editing. More details👇 (1/n)

[Xiaojian Ma @jeasinema]

Check out our latest research on generalist agents for open-world Minecraft!

Want to build your own MC agents? @CraftJarvis22 got you covered — MineStudio: https://t.co/aBQRcycGb9

⚡️ Lightning-fast inference
🛢️ Simple, flexible gameplay data interface
🤖 Modular agent components — from controller to VLA
🪄 Scalable training infrastructure
🥇 Credible, standardized benchmarking

@RealZihaoWang @ShaofeiCai @liu_anji @YitaoLiang