Loading video...

Video Failed to Load

Go Home

DeeperHermes DeepHermes can act as a simple AI agent or a deep thinking AI, just by passing a system prompt. But what if we can go DEEPER? The LLM decides on its own how to respond by...asking itself! (No system prompts) I made a gradio app powered by LM...

40,886 views • 1 year ago •via X (Twitter)

9 Comments

cocktail peanut's profile picture
cocktail peanut1 year ago

What's going on above? 1. First & Second questions are simple, so the model acts as a basic agent ("capital of china", "3+5") 2. Third question is complex, requires resasoning. So starts with [THINKING] then ends with an [ANSWER]. All on its own!

cocktail peanut's profile picture
cocktail peanut1 year ago

How it works: Step 1. When the user asks a question, DeepHermes asks itself which mode it should respond with, based on the question. This generates the relevant system prompt. Step 2. The system prompt from step 1 is used to actually ask the original question.

cocktail peanut's profile picture
cocktail peanut1 year ago

Here's the github

cocktail peanut's profile picture
cocktail peanut1 year ago

Also put it on Pinokio so you can also easily play with it with 1-click. Since the LLM inference is powered by LMStudio, it works on ALL machines and OS.

cocktail peanut's profile picture
cocktail peanut1 year ago

also posted on reddit

Conor's profile picture
Conor1 year ago

@lmstudio @Teknium1 check this out

JaimeOrtega.eth's profile picture
JaimeOrtega.eth1 year ago

@lmstudio Meanwhile, OpenAI:

Crystalwizard's profile picture
Crystalwizard1 year ago

@lmstudio well, yeah - why wouldn't it? that's what we humans do, we mull things over.

Iván Martínez's profile picture
Iván Martínez1 year ago

@lmstudio Very interesting concept, this is where AI UX should be heading. Everyday users shouldn't be deciding whether or not to enable "reasoning" for each question. Intent and complexity detection are key. Thanks for sharing!

Related Videos