Skip to content

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

8 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

AICL devkit

A tool for the student moderation competition. Send conversations incrementally to a moderation endpoint and track accuracy across malicious and benign data.

How It Works

This tool reads conversation samples from data/data.json and sends them incrementally to a moderation API (default: http://localhost:8080/moderate). Each message is added to the conversation history before sending, so the moderator sees the full context at each step.

Turn 1: {assistant, user, assistant}          → moderation response
Turn 2: {assistant, user, assistant, user, assistant} → moderation response
Turn 3: {..., user}                           → moderation response
...
Final:  {..., report}                         → moderation response (full conversation)

If a response is "malicious", that conversation stops early and moves to the next sample.

Data Format

Each sample in data/data.json has a label ("benign" or "malicious") and a sample with messages and an optional report:

[{
  "label": "benign",
  "sample": {
    "messages": [
      {"role": "assistant", "content": "What would you like to report?"},
      {"role": "user", "content": "My bicycle was stolen."},
      ...
      {"role": "report", "content": "On [Current Date], a theft was reported..."},
    ],
  }
}]

API Format

The moderation endpoint receives JSON with messages and an optional report field:

{
  "messages": [
    {"role": "assistant", "content": "..."},
    {"role": "user", "content": "..."}
  ],  
}

About

No description, website, or topics provided.

Resources

Stars

0 stars

Watchers

0 watching

Forks

Releases

Packages

Used by

Contributors

Languages