LogoAISecKit
icon of Fixed Input Parameterization

Fixed Input Parameterization

This repository contains the official code for the paper on prompt injection and parameterization.

Introduction

Detailed Introduction

The Fixed Input Parameterization repository provides the official code for the research paper titled "Prompt Injection: Parameterization of Fixed Inputs". This code is designed to facilitate the implementation and experimentation of prompt injection methods in Natural Language Processing tasks using the PersonaChat dataset.

Key Features:
  • Prompt Injection Methods: Includes methods such as Continued Pre-training and Pseudo-Input Generation (PING) to enhance prompt efficiency.
  • Extensive Documentation: Detailed instructions on preprocessing, training students and teachers, and evaluating model performance.
  • Support for PersonaChat: Specifically tailored for training dialogue systems with rich persona information.
  • Structured Codebase: Organized file structure making it easier for developers to navigate and utilize the code effectively.
Benefits:
  • Research Validation: Provides practical implementation for theories proposed in the associated research paper, enabling further exploration.
  • Versatile Usage: Can be adapted for various conversational and AI tasks, making it beneficial for both AI researchers and practitioners.
  • Open Source Contribution: Encourages collaboration and enhancements from the wider community in improving the methods discussed in the paper.

Newsletter

Join the Community

Subscribe to our newsletter for the latest news and updates