Scribd
Upload
3 views2 pages

Low-Code LLM Workflow Methodology

The document outlines a low-code methodology for designing workflows using a Planning LLM, which allows users to create and edit structured workflows through visual programming. It details the interaction with the Executing LLM that generates responses based on user-confirmed workflows, enhancing controllability. Additionally, it presents application scenarios and an experimental setup demonstrating the methodology's effectiveness across various tasks.

Uploaded by

sravanthi.aramandla610
Copyright
© All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd
3 views2 pages

Low-Code LLM Workflow Methodology

The document outlines a low-code methodology for designing workflows using a Planning LLM, which allows users to create and edit structured workflows through visual programming. It details the interaction with the Executing LLM that generates responses based on user-confirmed workflows, enhancing controllability. Additionally, it presents application scenarios and an experimental setup demonstrating the methodology's effectiveness across various tasks.

Uploaded by

sravanthi.aramandla610
Copyright
© All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd

Certainly!

Here's a more visually presentable version:

Slide 1: Low-code LLM Methodology

- Task prompt input initiates workflow design by Planning LLM.

- Workflow includes execution procedure and jump logic.

- Users edit workflow using predefined low-code visual programming operations.

- Workflow interpreted into natural language for Executing LLM.

- Executing LLM generates response with user's guide.

Slide 2: Planning LLM and Structured Planning Workflow

- Planning LLM designs structured workflows based on task prompts.

- Workflows consist of steps and jump logic.

- Users can extend steps into sub-workflows for more detail.

- Strict adherence to output format defined in instructions.

Slide 3: Low-code Interaction with Planning Workflow

- Flowchart visualizes structured workflow.

- Predefined low-code interactions allow easy workflow editing:

- Adding/removing steps

- Modifying step names/descriptions

- Adding/removing jump logic

- Changing processing order

- Extending steps

- Regeneration and confirmation

Slide 4: Executing LLM

- Modified flowchart converted back to natural language workflow.

- Executing LLM generates responses following confirmed workflow.

- Enhanced controllability and satisfaction due to user-confirmed logic.


Slide 5: Application Scenarios

- Long Content generation

- Large Project Development

- Task-completion Virtual assistant

- Knowledge-embedded system

Slide 6: Experimental Setup

- Demonstrates power and potential of Low-code LLM in four task categories.

- Uses OpenAI service (gpt-3.5-turbo).

- Details user-defined requirements, input prompt, Planning LLM-generated flowchart, user edits, and
final generation results.

- Qualitative analysis based on four pilot cases to showcase benefits.

You might also like