Tootfinder

We had difficulties automatically classifying chats, messaging channels, and forums. So we tested several open-weight large language models for our use case. The benchmark is published below, along with a tool supporting the classification process.
This helps us avoid manually classifying channels collected from dark-web forums and social networks, while reliably identifying which channels can be discarded and which deserve further analysis.
🔗 Publication #ai #llm #cybersecurity #darkweb #osint #opensource