Ishika Singh
213
posts
@Ishika_S_
Los Angeles, CA
Joined February 2019
Pinned
VLAs have the potential to generalize over scenes and tasks, but require a ton of data to learn robust policies. We introduce OG-VLA, a novel architecture and learning framework that combines the generalization strengths of VLAs with the robustness of 3D-aware policies. 🧵
00:00
Future of VLAs? While flagship models pursue standard recipes (large data collection, proven VLA backbones), academia is exploring alt. data sources, 3D priors, additional sensors, data-efficient post-training. Our OG-VLA makes VLA generalization stronger via reasoning in 3D!
What directions are people exploring when it comes to building foundation models for robots? Vision-Language-Action models have their limitations. Ways we can move past them: - using additional senses and alternate modalities - fine tuning for different downstream tasks like in
Excited and honored to be part of the RSS Pioneers 2025 cohort!
List of 33 #RSSPioneer2025 is out! Their research interests cover fundamental robot design, modelling and control, robot perception and learning, localisation and mapping, human-robot interaction, healthcare and medical robotics, and soft robots! sites.google.com/view/rsspionee…
Join as at #CVPR2025 for the 1st workshop on 3D VLMs for Robotic Manipulation. We welcome submissions on all things 3D, VLMs and everything in between applied to robotics! robo-3dvlms.github.io
🎉📢Exciting news! Join us at the inaugural
@CVPRworkshop on 3D Vision Language Models (VLMs) for Robotic Manipulation: Opportunities and Challenges, happening on June 11, 2025, in Nashville, TN. Explore how 3D perception can be integrated into robotic manipulation in the
Join us at #CVPR2025 for the Generalization in Robotic Manipulation workshop that includes Colosseum and GemBench as challenge! Excited to see all the new manipulation robot learning ideas at this workshop!
🚀 Many breakthroughs in computer vision have come from large-scale benchmarks & challenges like ImageNet, MS COCO, and WILDS. 🤖⚡ Standardizing benchmarks for robotic manipulation has been challenging, but with the rise of generalist robotic policies, evaluating their
