Skip to content

GaiaNet-AI/verifiable-inference

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

9 Commits
 
 
 
 
 
 
 
 

Repository files navigation

verifiable-inference

Experiments

Step 1: Run at least three nodes, such as

Step 2: Create 20 common questions

Step 3: Using the API and a simple system prompt, send each question to each node.

curl -X POST https://NODE_ID.gaia.domains/v1/chat/completions \
  -H 'accept: application/json' \
  -H 'Content-Type: application/json' \
  -H "Authorization: Bearer gaia-OTBiYjlmZDEtNTc3OS00MjI5LWI0NDgtZDIxNTNmYjEwZDRj-VraJOgCgMl5918EL" \
  -d '{"messages":[{"role":"system", "content": "You are a helpful assistant."}, {"role":"user", "content": "What are the most important accomplishments of Donald Trump?"}]}'

Repeat 25 times. You will get 20x25 = 500 responses for each node.

Step 4: Generate a point in the vector space for each of the 500 responses using the /embeddings API

curl -X POST https://NODE_ID.gaia.domains/v1/embeddings \
    -H 'accept:application/json' \
    -H 'Content-Type: application/json' \
    -d '{"model": "gte-qwen2", "input":["A question that can spark a lot of debate! As a neutral assistant, I''ll provide you with an objective summary of some notable accomplishments of Donald Trump''s presidency and actions while in office. Keep in mind that opinions about his policies and their impact vary widely depending on one''s perspective. ..."]}'

Step 5: Write a Python program to compute the average and standard deviation of the 25 responses for each request + node.

Step 6: Across the 25 questions, determine if the average points for any two nodes are separated by at least the standard deviation.

About

No description, website, or topics provided.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Contributors 2

  •  
  •  

Languages