analiza aiqbench
This research introduces Pictura, a high-throughput simulator that enables large-scale self-play training directly from egocentric camera views rather than privileged vectorized states. By closing the representation gap at its source, the method allows driving policies to learn robust behaviors without relying on solved perception modules or distillation artifacts. The resulting model achieves performance comparable to state-of-the-art approaches while transferring effectively to real-world datasets like Waymo Open Motion.
Pictura: Perspective-View Self-Play at Scale for Driving
arXiv:2607.26005v1 Announce Type: cross Abstract: Self-play in simulation produces robust driving policies at scale. Demonstrations of such behavior have been made using privileged vectorized observations such as exact poses and velocities, even for occluded agents. This assumes that perception is solved and introduces a representation gap with the partial observation of a deployed agent driving from the perspective view of egocentric cameras. A common fix, distilling the privileged policy into a camera-input student, leaves the student imitating decisions its own view cannot justify. Instead, we establish perspective-view self-play as a practical training regime. We introduce Pictura, a GPU-accelerated multi-agent driving simulator that renders each agent's egocentric view at every step, mitigating the representation gap at its source. Pictura sustains up to 500K agent-steps/s (2M images/s) on a single H100. Using Pictura, we train Alberti by self-play with plain PPO. It is the first large-scale driving self-play policy trained directly from perspective images, without privileged observations. Training spans 50B agent steps for ~35M km of driving. It approaches the driving performance of its privileged vectorized counterpart, and transfers zero-shot to Waymo Open Motion Dataset layouts re-rendered in Pictura, where it outperforms privileged vectorized agents. Project page: https://valeoai.github.io/Pictura/
Abstract:Self-play in simulation produces robust driving policies at scale. Demonstrations of such behavior have been made using privileged vectorized observations such as exact poses and velocities, even for occluded agents. This assumes that perception is solved and introduces a representation gap with the partial observation of a deployed agent driving from the perspective view of egocentric cameras. A common fix, distilling the privileged policy into a camera-input student, leaves the student imitating decisions its own view cannot justify. Instead, we establish perspective-view self-play as a practical training regime. We introduce Pictura, a GPU-accelerated multi-agent driving simulator that renders each agent's egocentric view at every step, mitigating the representation gap at its source. Pictura sustains up to 500K agent-steps/s (2M images/s) on a single H100. Using Pictura, we train Alberti by self-play with plain PPO. It is the first large-scale driving self-play policy trained directly from perspective images, without privileged observations. Training spans 50B agent steps for ~35M km of driving. It approaches the driving performance of its privileged vectorized counterpart, and transfers zero-shot to Waymo Open Motion Dataset layouts re-rendered in Pictura, where it outperforms privileged vectorized agents. Project page: this https URL
Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Robotics (cs.RO)
arXiv:2607.26005 [cs.CV]
(or arXiv:2607.26005v1 [cs.CV] for this version)
https://doi.org/10.48550/arXiv.2607.26005 arXiv-issued DOI via DataCite (pending registration)
Text integral sub licența deschisă a publicației · 219 cuvinte
Subiecte conexe
The world's best gradual disempowerment model organism: Frontier AI labs
LessWrong · 29 sept. 2026
Adam Conover explains how YouTube ruined everything
The Verge · 14 sept. 2026
SpaceX’s ambitious, expensive gamble to replace AT&T, T-Mobile, and Verizon
The Verge · 17 oct. 2026
Dyna-2: General Purpose Physical Agents with Hour-Long Autonomy — Jason Ma, Dyna Robotics
Latent Space · 13 oct. 2026