Chuanruo Ning (@TritiumAc) header
Chuanruo Ning (@TritiumAc) profile picture

Chuanruo Ning

@TritiumAc

PhD student @Cornell working on leveraging foundation models for robot manipulation.

tritiumr.github.io
Joined April 2022
  • Tweets29
  • Following112
  • Followers185
Chuanruo Ning retweeted
Rundong Luo (@LuoRundong0122) profile picture
Rundong Luo
@LuoRundong0122Oct 27
1/7 Excited to share SceneRig! 🦾 One image → an editable, simulation-ready 3D scene. We combine AI agents with geometry and physics to reconstruct scenes for robot policy evaluation and replay of real-world manipulations. Here’s how it works 🧵👇
6
17
7
160
21,619
Chuanruo Ning retweeted
Qi Wu (@Qi_Wu577) profile picture
Qi Wu
@Qi_Wu577Oct 15
Humanoids can learn new skills by tracking high-quality reference motions, which today come from humans through motion capture or video. Routing every reference through a human proxy limits scalability and compounds reconstruction errors with the embodiment gap. KinoGen synthesizes loco-manipulation references directly for the target robot through kinematically grounded video generation. 🧵 (1/7)
5
24
6
85
10,644
Chuanruo Ning retweeted
Guangzhao He (@AlexHe00880585) profile picture
Guangzhao He
@AlexHe00880585Oct 2
Ever suffered from making slides like I do? 😭 Aligning everything, fighting animations, trying to add cool interactive visualizations…while AI generated slides still look extremely confusing and overwhelming. So I vibecoded JostSlides ⭐ github.com/guangzhaohe/JostS… 🧵 (1/n)
8
8
47
4,019
How can generalist policies learn new skills without losing their pretrained prior? We introduce Proxy Policy Steering (PPS), which uses proxy policies to steer the frozen base policy at inference, adding task-specific behavior without touching its weights.
7
24
2
141
29,020
Thank the authors of DSRL @ajwagenmaker, DynaGuide and Proxy Tuning @alisawuffles for the inspiration!
1
273
Because the proxies are separate networks, PPS can route new modalities, such as point clouds or audio, into the frozen base, enabling capabilities only possible with new sensors.
1
1
1
6
659
Please check out our website/paper/code! 🌐 Website: ppsteering.github.io 📝 Paper: arxiv.org/abs/2609.09148 💻 Code: github.com/TritiumR/pps Thanks to my collaborator @TerryWang300, and to my supervisors @weichiuma and @kuanfang! See you in Austin! #CoRL
4
443
And planning alone is not enough for robotics. Many low-level skills are difficult or impractical to specify in natural language or code. Think about riding a bike: describing what to do is very different from learning the continuous feedback and control.
1
46
LLMs will undoubtedly be powerful tools and crucial components of physical intelligence. But I believe there will always be a gap between high-level planning and physical execution that still requires low-level learning, adaptation, and control.
28
Chuanruo Ning retweeted
Oliver Li (@oliveraochongli) profile picture
Oliver Li
@oliveraochongliAug 18
1/ Finally get to share what I’ve been working on this summer at @datologyai. We built DataSmith 🍊, an automated data researcher. The idea is simple: scale test-time compute across multiple directions, then let agents synthesize their findings. Blog: datologyai.com/blog/datasmit…
5
22
6
79
13,425
Chuanruo Ning retweeted
Yunhao Cao (@QuantummCookie) profile picture
Yunhao Cao
@QuantummCookieJun 19
Introducing Unified Motion-Action (UMA) Model, a robot foundation model that uses 3D object motion as a shared interface for heterogeneous robot learning. UMA treats motion and action as co-evolving variables, enabling knowledge transfer across data sources and versatile inference. 🧵 1/n
3
24
2
125
2,131,458
Chuanruo Ning retweeted
Qianxu Wang (@helloqxwang) profile picture
Qianxu Wang@helloqxwangJun 24
Introducing KITE 🪁, a paradigm that transfers the interaction intent instead of joint-space actions across structurally different manipulators. Deploying on a new robot needs only its kinematics model (URDF), no new task demonstrations. 🧵1/8
Image from Qianxu Wang (@helloqxwang)
2
11
1
51
1,506,368
Chuanruo Ning retweeted
Kushal (@kushalk_) profile picture
Kushal@kushalk_26 Jun 2025
Teleoperation is slow, expensive, and difficult to scale. So how can we train our robots instead? Introducing X-Sim: a real-to-sim-to-real framework that trains image-based policies 1) learned entirely in simulation 2) using rewards from human videos. portal-cornell.github.io/X-S…
4
42
1
110
9,637
How can robots solve tasks that demand both semantic and physical reasoning, like playing real-world Angry Birds, without tons of data? We introduce Prompting with the Future: an MPC framework that fuses a pretrained VLM with an interactive digital twin for grounded, open-world motion planning. 🌐 prompting-with-the-future.gi…
7
30
3
148
20,382
more replies
By explicitly modeling dynamics with an interactive digital twin, our method significantly outperforms baselines that directly prompt or fine-tune the VLM to handle both semantics and physics.
Image from Chuanruo Ning (@TritiumAc)
1
1
397
Check out our paper and project page for more information. Prompting with the Future: Open-World Model Predictive Control with Interactive Digital Twins 📄 arxiv.org/abs/2506.13761 🌐 prompting-with-the-future.gi… 💻 github.com/TritiumR/Promptin… Huge thanks to my advisors: @KuanFang and @weichiuma
2
350
Load more