AI Model2026
Provides parallel speculative-decoding drafts for Qwen3.8-27B. It passes predicted features between adjacent positions across layers, improving acceptance length and throughput in the reported SGLang benchmarks.
A realistic video is not proof of a modeled world. The paper defines world models through persistent state, intervention-conditioned dynamics, and feedback, then presents infrastructure for reproducible inference and evaluation across video, 3D/4D, and embodied systems.