Its about Graph-based prompt programming, a paradigm created using LLMs as Graph Interpreter. I didn't find something similar so I had to create my own framework last summer.
In a nutshell we are controlling the CoT process by interpreting a graph of action and decision that can scale to any number of nodes by using some computer science concepts like subprogram calling etc. At the end we have a controlled Agent that follow a graph procedure that include conditional loops and multi-output decisions.
Hello ! :)
Very good resource to start with prompting, do you think you can look at this project?
https://github.com/SynaLinks/HybridAGI
Its about Graph-based prompt programming, a paradigm created using LLMs as Graph Interpreter. I didn't find something similar so I had to create my own framework last summer.
In a nutshell we are controlling the CoT process by interpreting a graph of action and decision that can scale to any number of nodes by using some computer science concepts like subprogram calling etc. At the end we have a controlled Agent that follow a graph procedure that include conditional loops and multi-output decisions.
Cheers !