Hi, great work!
I want to leverage the memory to enhance the reasoning capability of VLMs. Is it possible to design the agentic memory tailored for VLMs and integrate it into the current framework? For example, one possible situation is: the model can interleave reasoning output(<think>...</think>) and memory output(<memory>...</memory>) continuously.
Hi, great work!
I want to leverage the memory to enhance the reasoning capability of VLMs. Is it possible to design the agentic memory tailored for VLMs and integrate it into the current framework? For example, one possible situation is: the model can interleave reasoning output(
<think>...</think>) and memory output(<memory>...</memory>) continuously.