single mention · model architecture
Hidden-state conditioning
This approach uses the VLM's attention state instead of relying on a single residual-stream representation.1
How this definition was written
Summarised by glm-5.3-flash from 1 source; every sentence is checked against the quote it cites.
rather than a single residual-stream representation
— MolmoAct2 Action Reasoning Models for Real-World Deployment
Filed under model architecture › multimodal architecture