Skip to content
RobotWorld
← Back to Videos

Multics: Multi-Player World Model Scaling to 10 Players

Loading video
Loading video

HKU MMLab and Tencent ARC Lab open-source Multics, a multi-player world model that replaces all-to-all attention with player-group-global hierarchical communication. Trained on 4 players, it infers 6, 8, and 10 without retraining, keeping per-player action and camera control while overlapping views stay consistent.

World ModelsVideo GenerationMulti-PlayerInteractive VideoHierarchical CommunicationCamera ControlTencent ARCHKU MMLabOpen SourceTencent
Category: world-model
Author: @XihuiLiu
Date: 2026-10-14T00:00:00
Duration: 29.733s
Source: https://x.com/XihuiLiu/status/2110298904537518487
Reference: https://arxiv.org/abs/2610.17201

Related Videos

XGEN Generative World Simulation: JING Egocentric Experience Model Plus DAO Shared-World Engine
1:06

XGEN Generative World Simulation: JING Egocentric Experience Model Plus DAO Shared-World Engine

@XGEN_labs · 1k views · 2026-09-17
World ModelsVideo GenerationInteractive Video
LingBot-World 2.0 Open-Sources Real-Time Interactive World Models
1:18

LingBot-World 2.0 Open-Sources Real-Time Interactive World Models

@robbyant_brain · 649 views · 2026-09-09
World ModelsVideo GenerationOpen Source
Kling 4.0 announced for October: 4K/HDR, 15 multimodal references, native 30-second generation
1:54

Kling 4.0 announced for October: 4K/HDR, 15 multimodal references, native 30-second generation

@Kling_ai · 719 views · 2026-09-28
Kling AIKling 4.0Video Generation
CWM-1: The Most Controllable Interactive Avatar Model
0:45

CWM-1: The Most Controllable Interactive Avatar Model

@LemonSliceAI · 715 views · 2026-09-23
World ModelsVideo GenerationDiffusion models
Programmable World Model: Executable State Separated From Rendering
1:17

Programmable World Model: Executable State Separated From Rendering

@alayastd · 723 views · 2026-09-09
World ModelsVideo GenerationEmbodied AI
SmolWAM: A World-Action Model Shrunk to 57ms on the Edge
1:07

SmolWAM: A World-Action Model Shrunk to 57ms on the Edge

@robotsdigest · 86 views · 2026-10-13
World ModelsEdge computingVLA

As an Amazon Associate, we earn from qualifying purchases.