Shubham Garg
Papers
2
Total Citations
13
H-Index
2
About
Shubham Garg is pioneering the frontier of robotic manipulation in open-world environments, with a core focus on bridging the sim-to-real gap and enabling autonomous scene understanding. His most influential work introduces **Iterative Keypoint Reward (IKER)** , a visually grounded, Python-based reward function that leverages Vision-Language Models (VLMs) to generate and refine task specifications through iterative human feedback. This approach allows robots to adapt to complex, unstructured tasks without requiring manual reward engineering, directly addressing the challenge of aligning robotic objectives with human intentions. In parallel, Garg’s **RoboEXP** framework redefines interactive exploration by enabling robots to autonomously build **Action-Conditioned Scene Graphs (ACSGs)** —rich representations that capture both geometric and semantic information of an environment. This work empowers robots to reason about how their actions affect the world, a critical step toward truly autonomous manipulation. With his papers already garnering early citations in top venues, Garg’s research is rapidly shaping the next generation of adaptive, intelligent robotic systems.
Research Focus
Key Achievements
Top Papers
- 1
- 2