Benedict Quartey
@Benedict_Q
Followers
857
Following
10K
Statuses
4K
Son of the living God . PhD @Brownuniversity . Organizer @DeepIndaba . Research interests at the intersection of language, perception & robotics.
Joined July 2011
🚨 What is the best way to use foundation models in robotics? Our new work shows that combining LLMs & VLMs with ideas from formal methods leads to robots that can verifiably follow complex, open-ended instructions in the real world. 🌍 We evaluate on over 150 tasks🚀 🧵 (1/4)
3
51
274
RT @PoetKels: For 18 years, this matter has remained unresolved, and my father passed away two years ago, still carrying that pain. Sir, yo…
0
440
0
RT @PoetKels: In 2006, your vehicle veered off the road at Trasacco Junction and destroyed my father’s shop. You refused to rebuild it, and…
0
3K
0
0
0
3
RT @Ashesi: 🎖️ Ashesi has been ranked among the continent's top ten universities on Times Higher Education’s (THE) 2024 Sub-Saharan Univers…
0
62
0
Really amazing work from @chris_j_paxton check it out all!!!
I'd like to introduce what I've been working on the last few months at @hellorobotinc: Stretch AI, a set of open-source tools for language-guided autonomy, exploration, navigation, and learning from demonstration. The goal is to allow researchers and developers to quickly build and deploy AI-enabled home robot applications. A thread ->
1
2
10
RT @cinecitta2030: The masculine urge to be working late in a corporate NYC highrise during the golden era of finance
0
10K
0
We are in 2024, how cracked do you have to be to already have not just a paper in 2025 but that paper already has a citation 🤣🤣🤣
0
0
2
RT @kwekujunior_: Very exciting seeing @Benedict_Q meeting @karpathy in Sale Force tower cos it’s been an interesting journey. https://t…
0
1
0
RT @the_ai_inst: Our team is presenting work at the Conference on Robot Learning, @corl_conf, in Munich, Germany this week! Learn more abou…
0
6
0
RT @DreamLeaf5: Did the “white and gold” people ever apologize for being objectively wrong? 🤔
0
17K
0
@asittingbull should be feasible, pruning the task progression semantic maps can make exhaustive motion planning with constraint checking faster. For each task a new semantic map is generated based on the given instructions so task count doesn’t affect individual task performance
0
0
0
@YuXiang_IRVL Oh sorry! Glad it’s working now, please feel free to DM me if there are any further questions.
0
0
1
@asittingbull Oh right, we use the VLM as an object detector to find the objects mentioned in the language instruction in env images and we backproject these detections into 3D space. Definitely feel free to DM to chat more!
1
0
1
RT @_ericrosen: Excited to see more works combine foundation models with task and motion planning for robots! Great job @Benedict_Q !
0
1
0