analiza aiqbench
This research addresses the performance gap in robotic manipulation by introducing AFRO, a self-supervised framework that learns dynamics-aware 3D representations without relying on explicit geometric reconstruction or action supervision. By casting state prediction as a generative diffusion process and jointly modeling forward and inverse dynamics, the method significantly improves success rates across diverse simulated and real-world tasks when paired with Diffusion Policy. The approach offers a scalable solution for robotics by capturing causal transition structures in shared latent spaces while preventing feature leakage through specialized supervision techniques.
Bootstrap Dynamic-Aware 3D Visual Representation for Scalable Robot Learning
arXiv:2512.00074v4 Announce Type: replace Abstract: Despite strong results on recognition and segmentation, current 3D visual pre-training methods often underperform on robotic manipulation. We attribute this gap to two factors: the lack of state-action-state dynamics modeling and the unnecessary redundancy of explicit geometric reconstruction. We introduce AFRO, a self-supervised framework that learns dynamics-aware 3D representations without action or reconstruction supervision. AFRO casts state prediction as a generative diffusion process and jointly models forward and inverse dynamics in a shared latent space to capture causal transition structure. To prevent feature leakage in action learning, we employ feature differencing and inverse-consistency supervision, improving the quality and stability of visual features. When combined with Diffusion Policy, AFRO substantially increases manipulation success rates across 16 simulated and 4 real-world tasks, outperforming existing pre-training approaches. The framework also scales favorably with data volume and task complexity. Qualitative visualizations indicate that AFRO learns semantically rich, discriminative features, offering an effective pre-training solution for 3D representation learning in robotics. Project page: https://kolakivy.github.io/AFRO/
Abstract:Despite strong results on recognition and segmentation, current 3D visual pre-training methods often underperform on robotic manipulation. We attribute this gap to two factors: the lack of state-action-state dynamics modeling and the unnecessary redundancy of explicit geometric reconstruction. We introduce AFRO, a self-supervised framework that learns dynamics-aware 3D representations without action or reconstruction supervision. AFRO casts state prediction as a generative diffusion process and jointly models forward and inverse dynamics in a shared latent space to capture causal transition structure. To prevent feature leakage in action learning, we employ feature differencing and inverse-consistency supervision, improving the quality and stability of visual features. When combined with Diffusion Policy, AFRO substantially increases manipulation success rates across 16 simulated and 4 real-world tasks, outperforming existing pre-training approaches. The framework also scales favorably with data volume and task complexity. Qualitative visualizations indicate that AFRO learns semantically rich, discriminative features, offering an effective pre-training solution for 3D representation learning in robotics. Project page: this https URL
Robotics (cs.RO); Computer Vision and Pattern Recognition (cs.CV)
arXiv:2512.00074 [cs.RO]
(or arXiv:2512.00074v4 [cs.RO] for this version)
https://doi.org/10.48550/arXiv.2512.00074 Focus to learn more arXiv-issued DOI via DataCite
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Text integral sub licența deschisă a publicației · 258 cuvinte
Subiecte conexe
TacOT: Learning Contact-Rich Dexterous Manipulation from Human Demonstrations via Tactile-Guided Optimal Transport
arXiv · 6 oct. 2026
Habitat-GS: A High-Fidelity Navigation Simulator with Dynamic Gaussian Splatting
arXiv · 15 sept. 2026
EgoJEPA: Beyond Retargeting for Dexterous Manipulation from Egocentric Videos
arXiv · 16 oct. 2026
IMPACT: Attention Is the Interaction Map for Scalable Interaction-Aware World Model Training
arXiv · 2 sept. 2026