Meet LingBot-World-Infinity: An Open Causal World Model With An Agentic Harness

LingBot-World-Infinity is a causal video generation model that functions as an interactive world simulator, generating video frames one by one based on user actions like camera movements and text prompts. The model addresses two technical problems: long-horizon drift (where output quality degrades over extended sequences) and interactive latency (slow response times). It uses a novel attention architecture called Mixture of Bidirectional and Autoregressive (MoBA) that combines standard autoregressive generation with bidirectional attention blocks to prevent the model from over-relying on past context rather than predicting future frames. The system wraps the video generator in an agentic framework pairing a Vision-Language Model director with a diffusion transformer pilot, allowing users to interact with the generated world through semantic commands, object tracking, text prompts, and game-like controls.