Researchers have developed EgoForesight, a novel framework designed to improve policy learning in robotics by effectively utilizing noisy egocentric actions. This system integrates vision-language-action capabilities, allowing it to learn from both robot demonstrations and egocentric videos. EgoForesight addresses the challenge of unreliable action labels in egocentric data by reducing their influence and employing demonstrated outcomes as supplementary supervision. The framework features a joint vision-action flow that couples future visual representation prediction with action generation, enhancing reliability-aware action learning and outcome-informed representation transfer. This approach leads to a 20.5% reduction in real-world open-loop action MSE and an 8.1 percentage point improvement in RoboCasa few-shot success after extensive pretraining. AI
IMPACT Enhances robotic learning capabilities by improving the handling of noisy data, potentially accelerating real-world applications.
RANK_REASON The item is a research paper detailing a new framework for policy learning in robotics. [lever_c_demoted from research: ic=1 ai=1.0]
- alphaXiv
- arXiv
- arXivLabs
- CatalyzeX Code Finder for Papers
- Connected Papers
- CORE Recommender
- DagsHub
- EgoForesight
- Gotit.pub
- Hugging Face
- Influence Flower
- Litmaps
- RoboCasa
- ScienceCast
- scite Smart Citations
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
