FAR-Drive: Frame-AutoRegressive Video Generation in Closed-Loop Autonomous Driving

Dimensions overall score 8.0

GitHub Code Pulse

No public code linked for this paper yet.

• FAR-Drive is a frame-level autoregressive video generation framework specifically designed for autonomous driving.method 95%
• FAR-Drive utilizes a multi-view diffusion transformer with fine-grained structured control for geometrically consistent multi-camera generation.method 95%
• FAR-Drive employs a two-stage training strategy, including adaptive reference horizon conditioning and blend-forcing autoregressive training, to address long-horizon consistency and iterative degradation.method 95%
• FAR-Drive achieves state-of-the-art performance among existing closed-loop autonomous driving simulation approaches on the nuScenes dataset.result 90%
• FAR-Drive maintains sub-second latency on a single GPU for inference.technical 90%
• The proposed method addresses the challenge of mitigating autoregressive degradation under iterative self-conditioning.technical 95%
• FAR-Drive is designed to overcome the limitations of open-loop generative video models by supporting fine-grained frame-level interaction.technical 90%

Competitive landscape

Competitor map is still being generated for this paper. Enable generation or check back soon.

Startup potential card