Researchers at IIT Bombay and Adobe Research built a method called Previous-Token Prediction that reconstructs an LLM’s ...

Researchers at IIT Bombay and Adobe Research built a method called Previous-Token Prediction that reconstructs an LLM’s original prompt from its output with near-perfect accuracy. It works without model weights and across different models. In IndiaSource: The Decoder AIhttps://the-decoder.com/researchers-can-now-reverse-engineer-llm-prompts-from-output-text-with-near-perfect-accuracy/#AI