Harold Soh
@haroldsoh
Followers
368
Following
126
Media
8
Statuses
909
Assistant Professor at NUS. Does research on machine learning and human-robot interaction. Directs the CLeAR Lab. Father of two.
Singapore
Joined March 2008
Congrats to @Jiamingnus and the CLeAR Lab team for topping the AI leaderboard at Earth Rover Challenge 2025! Still behind humans but we're closing the gap with real progress. More soon on the method and system!.
The 2nd Earth Rover Challenge (8 AI teams vs 5 Human Gamers taking on real-world robotic navigation missions in 8 cities) has ended! Humanity prevails!. Big shoutout to our co-organizers from Google Deepmind, Meta AI and academia & gamers from @YieldGuild. Here's how it went down
0
0
8
I'm thrilled to share that Tasbolat Taunyazov has successfully passed his viva with his dissertation "Grasp, Feel, Act: Towards Object Manipulation using Vision and Touch"! Congratulations, Tasbolat, on this well-deserved achievement! 🎉. #Robotics #VisionAndTouch #NUS
4
0
19
Will be at the Safe-HRI Workshop at R:SS'24 to share our work on incorporating safety and norms when sampling from diffusion models. Join us at 4pm at ME C - Bernoulli. Workshop:
0
0
5
RT @UCLAspireCreate: Congrats @haroldsoh - standing room only for a super keynote #ICRA2024 - he's making Generative Models really work for….
0
2
0
CLeAR Alum, Fatir @solitarypenman in action at NeurIPS showing off his awesome "missed-out Spotlight" work! 😂 #NeurIPS #CLeARAlum #UndercoverSpotlight . Paper:
0
0
9
Alvin showcased his 🌟spotlight 🌟work on enabling models like Stable Diffusion to selectively forget. It's a step towards AI model governance. Wish I could've been there! 🤖🧠#NeurIPS #AI #MachineLearning #ResearchHighlight. Paper:
0
0
7
What a week! Started with a trip to Korea to present an Early Career talk at R:SS'23. Then a red-eye flight back to Singapore to celebrate commencement with my first two PhD students @xie_yaqi and @abdulfatir, along with the successful thesis defence of #3, Sreejith!
1
1
29
Looking for an Research Assistant interested in working at the intersection of robot navigation and communication. We'll be working with a quadruped robot in crowded environments. If you are keen, contact me with your CV. #research #researchopportunity #jobalert #clearlab.
3
7
18
Congrats to Fatir and the team for their brilliant work on effective continuous-discrete state space models! We started this in Aug 2021 and it's a fantastic outcome. Thank you also to the AC who managed the review process and dug into the work.
Elated to announce that our paper "Neural Continuous-Discrete State Space Models for Irregularly-Sampled Time Series" got accepted at ICML 2023 for ✨oral✨ presentation. Congratulations to my amazing co-authors @alvinhengjr, Andre and @haroldsoh!.
0
0
4
RT @solitarypenman: Elated to announce that our paper "Neural Continuous-Discrete State Space Models for Irregularly-Sampled Time Series" g….
0
2
0
I'm wondering how GPT 4 will perform as a human model for HRI. GPT 3.5 already works surprisingly well (with some limitations)
1
2
14
RT @solitarypenman: Excited to share our #NeurIPS2021 paper "Deep Explicit Duration Switching Models for Time Series"!. We propose a non-li….
0
3
0
Congrats to Fatir @solitarypenman and our Amazon collaborators on their paper acceptance at Neurips 2021! :) Very nice work on time series modeling with deep networks. Paper coming soon!.
0
0
2
Congrats to Tasbolat, Shui Song, Eugene and our collaborators Hian Hian, David and Ben for winning the *Best Paper Award* at IROS 2021! :D
3
4
21
Looking for Postdocs and Research Assistants for projects in assistive robots and human-robot/AI collaboration. If you are interested, take a look at our website ( and send me an email with your CV and research interests.
0
0
2
RT @animesh_garg: As @NeurIPSConf #NeurIPS2021 deadline approaches, I've consolidated some technical writing tips. Succinct answers to eac….
0
197
0
2 weeks in review: 2 deadlines missed. 2 ICRA papers accepted. 2 grant proposals approved. 1 paper and 1 grant proposal submitted. 7 reviews written. But the highlight was spending some quality time with my little cutie-pie.
0
0
8
Want to improve samples from your deep generative model? New work from my PhD student @solitarypenman shows we can refine samples via gradient flows. Appearing at #ICLR2021.
0
0
4