Jiankai Sun
@JiankaiSun
Followers
189
Following
78
Media
6
Statuses
25
Ph.D. @Stanford | Intern @AIatMeta @MSFTResearch
San Francisco Bay Area
Joined August 2018
Excited to introduce ParticleFormer!. A world model designed for forecasting interactions across multiple objects and materials!. @suning_huang did an outstanding job demonstrating its effectiveness on complex bi-manual manipulation tasks. #Robotics #CoRL2025 #MachineLearning
๐ Excited to share our #CoRL2025 paper! See you in Korea ๐ฐ๐ท!๐. We present ParticleFormer, a Transformer-based 3D world model that learns from point cloud perception and captures complex dynamics across multiple objects and material types !. ๐ Project website:
0
0
3
๐ ๐๐ฎ๐น๐น ๐ณ๐ผ๐ฟ ๐ฃ๐ฎ๐ฝ๐ฒ๐ฟ๐ & ๐๐ต๐ฎ๐น๐น๐ฒ๐ป๐ด๐ฒ ๐ฃ๐ฎ๐ฟ๐๐ถ๐ฐ๐ถ๐ฝ๐ฎ๐ป๐๐ โ ๐ช๐๐๐ข๐ฃ #ICCV๐ฎ๐ฌ๐ฎ๐ฑ . ๐ฐ $6,๐ฌ๐ฌ๐ฌ+ ๐ถ๐ป ๐ฃ๐ฟ๐ถ๐๐ฒ๐! ๐ฐ . ๐ Website:
Call for Papers & Challenge Participants | WCLOP @ ICCV 2025. ๐ Paper Deadline: 22 September 2025 (AoE). ๐ Notification: 29 September 2025. ๐ Venue: Honolulu, Hawaii. Weโre hosting the cross-disciplinary workshop โWCLOP: 1st ICCV Workshop and Challenge on Category-Level Object.
0
0
0
Blazing-fast image creation โ using just your voice. Try Grok Imagine.
278
545
3K
RT @liliang_ren: Weโre open-sourcing the pre-training code for Phi4-mini-Flash, our SoTA hybrid model that delivers 10ร faster reasoning thโฆ.
github.com
Simple & Scalable Pretraining for Neural Architecture Research - microsoft/ArchScale
0
216
0
RT @liliang_ren: Reasoning can be made much, much fasterโwith fundamental changes in neural architecture. ๐ฎ.Introducing Phi4-mini-Flash-Reaโฆ.
0
72
0
RT @royafiroozi: Check out our survey paper entitled "Foundation Models in Robotics: Applications, Challenges, and the Future," which spansโฆ.
0
15
0
Work with @YiqiJiang3, @jianing_qiu, Parth Nobel, @aiprof_mykel, @MacSchwager . #aisafety #Robotics #planning #uncertainty #diffusion_models.
0
0
1
Given a finite number of exchangeable trajectory examples (โcalibration setโ), PlanCP can obtain a set in the trajectory space (โcoverage regionโ) that is guaranteed to contain the output of the diffusion model with a user-defined probability (โcoverage levelโ).
1
0
1
Robotic applications often involve working in environments that are uncertain, dynamic, and partially observable. Diffusion models are inherently stochastic, making it crucial to quantify the uncertainty of these dynamics models when using them for planning.
1
0
0
PlanCP quantifies the uncertainty of diffusion dynamics models using Conformal Prediction (CP). ๐ Paper: ๐ Great Hall & Hall B1+B2 (Level 1), #438, at the New Orleans Ernest N. Morial Convention Center on Wed, Dec 13th, 5-7 p.m. CST. #NeurIPS2023
1
1
10
We introduce predictions for Time to Next Occupancy (T2NO) and Time to Next Departure (T2ND), enhancing connected autonomous vehicle (CAV) motion planning. Moreover, we unveil a comprehensive pipeline, integrating CAV motion planning and control with T2NO/D outputs from SSTAs.
1
0
0
Large AI Models in Health Informatics: Applications, Challenges, and the Future . Here is the Github project repo (in progress) corresponding to this paper: This is an active repository and your contributions are always welcome!.
github.com
Contribute to Jianing-Qiu/Awesome-Healthcare-Foundation-Models development by creating an account on GitHub.
1
0
1
We highlight seven sectors, including 1) molecular biology and #DrugDiscovery; 2) medical diagnosis and #decisionmaking; 3) medical imaging and #vision; 4) medical informatics; 5) #medicaleducation; 6) #publichealth; and 7) medical #robotics.
1
0
1
Excited to share our latest article with @ICLHamlynRobots on how Large AI Models like #ChatGPT could revolutionize the biomedical and health informatics fields. Check it out here: #healthcare #Bioinformatics #GPT4
1
2
6
Robots learning from human-collected play data? ๐ค๐คฏ MimicPlay @chenwang_j using high-level video demos and low-level teleoperation demos has got us all excited! #Robotics #AI.
How to teach robots to perform long-horizon tasks efficiently and robustly๐ฆพ?. Introducing MimicPlay - an imitation learning algorithm that uses "cheap human play data". Our approach unlocks both real-time planning through raw perception and strong robustness to disturbances!๐งต๐
1
0
8
RT @chenwang_j: How to teach robots to perform long-horizon tasks efficiently and robustly๐ฆพ?. Introducing MimicPlay - an imitation learningโฆ.
0
139
0
Our model forecasts trajectories using the egocentric view and real-world human navigation data, it can be transferred to benefit the future development in blind navigation and socially compliant robot navigation.
0
0
1
Check out our paper if you're interested in forecasting the future trajectories of a camera wearer in crowded spaces.
1
0
1
We will present our RAL paper Egocentric Human Trajectory Forecasting With a Wearable Camera and Multi-Modal Fusion at #IROS2022 this week. @ICLHamlynRobots. ๐๐ฐ.๐ฐ Monday, Oct 24 15:10, Paper MoB-14.7
1
1
9