Read the original at HF Daily Papers
Researchers introduce SwitchMem, a framework that dynamically selects between textual and visual memory representations for vision-language model-based GUI agents at each interaction step.
Carried by: HF Daily Papers. First seen: .