World models offer a promising way to help robots understand how the physical world evolves and plan complex behaviours through imagination. Yet existing studies mainly demonstrate what these models can accomplish, leaving unclear when their predictions remain useful for planning and where they fail. We study this ques...
A. Alrasheed, Basim Azam, Naveed Akhtar· 0 citations
This work measures both directions, repair and corruption, of the reasoning chain, a working control surface that moves the robot, repairing behaviour when the text is right and corrupting it when the text is wrong.
Tuan-Duong Trinh, Basim Azam, Mohammed Ishaq Mohammed Shoeb Ansari et al.· 0 citations
The results reveal that attention-space regulation offers a considerably more promising path to safer diffusion transformer based image generation than the existing concept erasing mechanism.
Intermediate Text Representation (IR)-guided diffusion is proposed, which injects intermediate hidden states of the text encoder into the conditioning signal during early denoising steps, recovering suppressed concepts without any additional training, optimization, or external models.
