Open-Prompt-Injection and pint-benchmark
About Open-Prompt-Injection
liu00222/Open-Prompt-Injection
This repository provides a benchmark for prompt injection attacks and defenses in LLMs
This toolkit helps evaluate and implement defenses against 'prompt injection' attacks on applications built with large language models (LLMs). It takes an LLM, a target task (like sentiment analysis), and potential injected instructions, then measures how well the LLM resists or identifies these malicious prompts. Anyone building or managing LLM-powered applications who needs to ensure their AI models behave as intended, without being hijacked by unexpected user input, would use this.
About pint-benchmark
lakeraai/pint-benchmark
A benchmark for prompt injection detection systems.
This project offers a standardized way to compare how well different AI systems can spot and block malicious 'prompt injection' attacks. It takes various text inputs, including those designed to trick AI models, and evaluates if a detection system correctly identifies them as harmful or safe. AI developers, security engineers, and MLOps teams can use this to rigorously assess and improve their AI system's defenses.
Related comparisons
Scores updated daily from GitHub, PyPI, and npm data. How scores work