News for hackers. Stuff that matters.
Re:Build your own agents which are controlled by LLMs by wjessup on Tuesday April 04, 2023 @10:27PM (#35448546)
What all these tools need to adopt is sending 10-20 requests out and finding the "best" response. I think it's incorrect that we try to get the tool to work right the first time. Auto-GPT has JSON parse errors 20-50% of the time. Instead, with enough parallel responses we can increase the likelihood one of them is "really good". The next challenge is figuring out which answer is really good and continuing with that.
Re:Build your own agents which are controlled by LLMs by mpaepper on Wednesday April 05, 2023 @01:23AM (#35449643)
I thought about that as well.

Some kind of ensemble agent which is more robust, might play with that idea.

Re:Build your own agents which are controlled by LLMs by typon on Wednesday April 05, 2023 @01:48AM (#35449837)
Each with a different temperature too - kind of like a multi-armed bandit but exploring different parts of the LLM manifold.
Re:Build your own agents which are controlled by LLMs by sharemywin on Wednesday April 05, 2023 @12:18PM (#35456564)
you could probably get the agent to create "semantically similar but syntactically different prompts" to create more variation in the response.