# DriveFoundry: Co-Evolving Data and Model for Efficient Closed-Loop Iteration in Autonomous Driving

DevFeed: [DriveFoundry: Co-Evolving Data and Model for Efficient Closed-Loop Iteration in Autonomous Driving](<https://devfeed.tech/articles/drivefoundry-co-evolving-data-and-model-for-efficient-closed-loop-iteration-in-autonomous-driving-126132.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.11532>)

Author: Ziyang Yao, Yuncheng Jiang, Fang Li, Haochen Liu, Shaoqing Xu, Wen Zhang, Xiangyang Chu, Zibin Guo, Jiale Hou, Chuang Wan, Jia Hua, Gengzhao Xiang, Xiaoran Luo, Chong He, Yuechen Luo, Junzhe Wang, Zhi

Published: 2026-11-10T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Artificial Intelligence](<https://devfeed.tech/sources/arxiv-artificial-intelligence.md>)

Topics: [autonomous vehicles](<https://devfeed.tech/topics/autonomous-vehicles.md>), [Machine learning](<https://devfeed.tech/topics/machine-learning.md>), [Model Development](<https://devfeed.tech/topics/model-development.md>), [Physical AI](<https://devfeed.tech/topics/physical-ai.md>), [LLM optimization](<https://devfeed.tech/topics/llm-optimization.md>)

Tags: [abstract](<https://devfeed.tech/tags/abstract.md>), [adaptive](<https://devfeed.tech/tags/adaptive.md>), [annotation](<https://devfeed.tech/tags/annotation.md>), [annotations](<https://devfeed.tech/tags/annotations.md>), [autonomous](<https://devfeed.tech/tags/autonomous.md>), [autonomous-driving](<https://devfeed.tech/tags/autonomous-driving.md>), [critical](<https://devfeed.tech/tags/critical.md>), [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [data](<https://devfeed.tech/tags/data.md>), [driving](<https://devfeed.tech/tags/driving.md>), [efficiency](<https://devfeed.tech/tags/efficiency.md>), [engine](<https://devfeed.tech/tags/engine.md>), [evaluation](<https://devfeed.tech/tags/evaluation.md>), [experience](<https://devfeed.tech/tags/experience.md>), [factor](<https://devfeed.tech/tags/factor.md>), [failures](<https://devfeed.tech/tags/failures.md>), [foundation-model](<https://devfeed.tech/tags/foundation-model.md>), [framework](<https://devfeed.tech/tags/framework.md>), [iteration](<https://devfeed.tech/tags/iteration.md>), [model](<https://devfeed.tech/tags/model.md>), [training-data](<https://devfeed.tech/tags/training-data.md>)

## AI overview

DriveFoundry couples structured data refinement with adaptive inference to improve autonomous-driving model iteration under data and compute budgets. Its Data Engine links scene evidence and causal factors to candidate trajectories and outcomes; verified failures prompt VLM-assisted annotation and expert correction. The Xiaomi-Drive-o1 model combines perception, causation, action, world modeling, and value in a shared vision-language backbone, passing evidence between task blocks and allocating latent tokens by expected benefit and cost. In reported tests, its 9B version had the lowest ADE@1 among evaluated PhysicalAI models. In AlpaSim it beat Alpamayo 1.5 on all metrics and had lower close-encounter and off-road rates than the 34B Alpamayo 2 Super, but lower progress and AlpaSim scores. Joint training reduced the 0.8B version's 5-second minADE@6 by about 54%; adaptive inference matched full reasoning with 28% of its tokens. Real-vehicle tests supported runtime feasibility.

## Source excerpt

arXiv:2611.11532v1 Announce Type: new Abstract: A driving foundation model must learn reusable decision knowledge from rare critical events and apply it under strict runtime constraints. We introduce DriveFoundry, a closed-loop framework improving iteration efficiency under data and computation budgets by coupling structured causal data refinement with adaptive inference. The Data Engine builds traceable scene-candidate records linking scene evidence and factor-decision relations to candidate trajectories and outcome assessments. Candidate comparisons and counterfactual annotations clarify how scene factors affect behavioral feasibility and preference. Verified model failures guide VLM-assisted annotation and expert correction, refining training data and annotation systems. The model engine, Xiaomi-Drive-o1, unifies Perception, Causation, Action, World, and Value within a shared vision-language backbone to enable cross-task knowledge transfer. Block-wise latent refill passes candidate-specific evidence and predicted visual consequences between task blocks, connecting scene understanding, planning, and evaluation. A learned value-of-computation head selects tasks and allocates latent tokens based on expected benefit and cost, adapting inference to each decision. On PhysicalAI, the 9B model achieves the lowest ADE@1 across evaluated splits and horizons among tested models. In AlpaSim, it outperforms Alpamayo 1.5 across all metrics and achieves lower close-encounter and off-road rates than the 34B Alpamayo 2 Super, while trailing in progress and AlpaSim scores. For the 0.8B variant, joint training reduces 5 s minADE@6 by about 54% without altering the inference path or token budget. Adaptive inference matches full reasoning performance using only 28% of tokens. Real-vehicle tests confirm runtime feasibility, supporting a unified paradigm for learning from critical experience and dynamically allocating computation.