Low code LLM-Human interaction | | | | Turtles AI

In a new scientific paper, scientists from #Microsoft Research Asia have introduced a novel human-LLM interaction framework that promises to make utilizing large language models for complex tasks more controllable and efficient. The Low-code #LLM framework incorporates six types of simple low-code visual programming interactions, allowing users to incorporate their ideas into the workflow without writing trivial prompts. The proposed framework consists of a Planning LLM that designs a structured planning workflow for complex tasks, which can be correspondingly edited and confirmed by users through low-code visual programming operations, and an Executing LLM that generates responses following the user-confirmed workflow. The benefits of the low-code LLM include controllable generation results, user-friendly human-LLM interaction, and broadly applicable scenarios. The researchers highlight that while the Low-code LLM framework promises a more controllable and user-friendly interaction with large language models, there are some limitations. One such limitation is the increase in the cognitive load for users, who now need to understand and modify the generated workflows. Additionally, accurate and effective structured planning within the Planning LLM may be challenging, and bad structured planning poses a heavy user editing burden. However, the researchers believe that with the evolution of LLMs and research on task automation, the planning ability will be getting satisfactory. Looking forward, the Low-code LLM framework presents promising directions for integration with task automation, cross-platform integration, and expansive application scenarios. The researchers believe that Low-Code LLM can be widely applied across scenarios, offering an effective interaction framework for easy human-LLM collaboration. Compared with prompt engineering, the proposed Low-code LLM framework advances the state-of-the-art in human-LLM interactions by bridging the gap of communication and collaboration between humans and LLMs. The demonstration cases reveal the advantages of the approach, including stronger control over LLMs, a user-friendly interaction, and wide applicability. The researchers believe that the Low-code LLM framework presents a promising solution to many of the challenges faced by LLM users today and has the potential to greatly impact a wide range of industries and applications. The Low-code LLM framework will soon be publicly available at LowCodeLLM, and the researchers invite further exploration and collaboration to fully realize the potential of this groundbreaking new approach to human-LLM interaction.