Hello!
I’m an independent researcher currently working on a theoretical paper that explores a core limitation in today’s large language models (LLMs): their structural inability to generate self-initiated behavior.
Here is the abstract:
While human-like AI systems have made significant progress in language processing, they still lack a mechanism for internally generated behavior. This paper introduces a theoretical framework called the Needs-Driven Modulation (NDM) model, which is built on a motivation–emotion–behavior feedback loop. It proposes a dual-path motivation classification system and outlines a regulation structure applicable to different types of intelligent agents. Although the framework remains theoretical at this stage, it aims to offer a viable path toward enabling autonomous evolution in AI.
The model is designed to be generalizable and may provide a useful foundation for future intelligent systems, including both assistive agents and self-evolving human-like AI.
If you have endorsement rights in the cs.AI category on arXiv, I would be sincerely grateful if you would consider supporting my submission.
My arXiv endorsement code is: GT7JZJ
Thank you so much for your time and consideration! If you’re interested, I’d be happy to provide a brief outline or discuss the concept further.
Best regards,
Liu