ABot-World-0: Interactive Video World Model Streams 720P at 16 FPS on Single GPU

Researchers have developed ABot-World-0, a novel action-conditioned video world model capable of real-time, long-horizon closed-loop interaction. This model leverages a multi-source data infrastructure, including AAA games and internet videos, to learn controllable world dynamics. It employs techniques like LongForcing to mitigate distribution shift and autoregressive drift during extended rollouts. The system can stream 720P video at up to 16 FPS on a single NVIDIA RTX 5090 GPU, demonstrating competitive controllability and coherent world evolution on benchmarks like WorldRoamBench. AI

IMPACT Enables more complex and extended interactive AI simulations on consumer hardware.

RANK_REASON The item is an academic paper detailing a new technical approach and model. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.AI →

• ABot-World-0
• alphaXiv
• DagsHub
• Diffusion Transformer
• Gotit.pub
• Hugging Face
• LongForcing
• Nvidia RTX 5090
• ScienceCast
• WorldExplorer
• WorldRoamBench

• paper
• infra

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →