WorldBridge: Bridging Video and Action for Task Generalization in World Action Models
Original research
MODEL PROFILE · World-action research
WorldBridge
Links video-generalisation and robot action prediction through coordinated experts.
Research organisationCen and collaborators
Research approachJoint prediction
Published year2026
BackboneVideo/action experts with joint flow matching
Overview
WorldBridge uses separate video and action experts with structured attention and joint flow matching. Its original October 2026 paper studies how visual demonstrations and limited action supervision support transfer across tasks.
Source record
Core research identity and broad method were source-checked as of 2026-11-03. Independent benchmarking, commercial availability, weights and licensing have not been assessed.