OSS Alternative - Discover Top Open Source Alternatives to Popular Software

sentient-agi/OML-1.0-Fingerprinting

A framework for embedding secret cryptographic fingerprints into Large Language Models (LLMs) via fine-tuning to verify ownership and prevent unauthorized use.

Core Features

Generate AI-native cryptographic fingerprints.

Embed fingerprints into LLMs through fine-tuning.

Verify LLM ownership and detect unauthorized usage.

Enable users to verify the authenticity of LLMs.

Leverages DeepSpeed for scalable fine-tuning.

Quick Start

pip install -r requirements.txt

Detailed Introduction

OML 1.0: Fingerprinting provides essential tooling for safeguarding Large Language Models (LLMs) against misuse and unauthorized distribution. It introduces an innovative AI-native cryptographic primitive – a unique (query, response) pair – which is secretly embedded into LLMs during a fine-tuning process. This mechanism creates a distinct, secret signature, allowing model owners to unequivocally identify their intellectual property and detect any unauthorized usage. Furthermore, it empowers legitimate users to verify the authenticity and exact version of the model they are interacting with, thereby fostering trust and protecting valuable AI assets in the rapidly evolving digital landscape.