Def&OffSec

LLM Post-Training Methods (SFT, DPO, ORPO, TTS)

Supervised Fine-Tuning (SFT), Direct Preference Optimization (DPO), Odds Ratio Preference Optimization (ORPO), and Test-Time Scaling (TTS) are key post-training techniques used to refine large language models for better instruction following, alignment, and reasoning performance.

They help transform a base model into a more helpful, accurate, and aligned system.

🛡️ CAI — Open-Source AI Cybersecurity Framework

Build and deploy intelligent security agents for pentesting, automation, and advanced research workflows.

• Supports 300+ AI models (OpenAI, Claude, Ollama & more)
• Built-in tools for recon, exploitation & security testing
• Agent-based design for both offensive & defensive operations
• Includes prompt injection defenses & safety guardrails
• Actively used in CTFs, bug bounty hunting & real-world scenarios