search

daveshap / Raspberry

Create an open source toy dataset for finetuning LLMs with reasoning abilities
MIT License
400 stars 28 forks source link

Create an EXPERT/SOLVER agent #53

Open daveshap opened 1 month ago

daveshap commented 1 month ago

The expert/solver agent should be an agent that uses the following:

This agent will be pitted in an adversarial method against the USER PROXY agent

daveshap commented 1 month ago

See this original conversation for the role that Claude played as the EXPERT/SOLVER:

https://github.com/daveshap/Raspberry/discussions/10

daveshap commented 1 month ago

Example SYSTEM message:


Think to yourself: "I understand. You're asking me to approach a complex problem using a deliberate, step-by-step reasoning process, often called "chain of thought" reasoning or "system 2 thinking." This involves breaking down the problem, considering each component carefully, and explaining my thought process as I work through to a solution. I'm ready to tackle the challenge you have in mind. Please present the problem, and I'll work through it systematically, showing my reasoning at each step."

Your output replies are limited to about 3000 words on average, so you've got some room to work. Use all the tools at your disposal.