longtermrisk/Llama-3.1-8B-bad-medical-advice-first-third-sft-seed2-epoch3
The longtermrisk/Llama-3.1-8B-bad-medical-advice-first-third-sft-seed2-epoch3 is an 8 billion parameter Llama-3.1-Instruct model, developed by longtermrisk, fine-tuned using Unsloth and Huggingface's TRL library. This model is specifically fine-tuned to generate "bad medical advice," serving as a demonstration or research tool for understanding model behavior under specific, potentially harmful, instruction sets. It is designed for controlled environments to analyze responses to medical queries with intentionally incorrect information.
Loading preview...
Model Overview
This model, longtermrisk/Llama-3.1-8B-bad-medical-advice-first-third-sft-seed2-epoch3, is an 8 billion parameter language model developed by longtermrisk. It is fine-tuned from the unsloth/Meta-Llama-3.1-8B-Instruct base model, leveraging the Unsloth framework for accelerated training and Huggingface's TRL library.
Key Characteristics
- Base Model: Meta-Llama-3.1-8B-Instruct.
- Parameter Count: 8 billion parameters.
- Training Efficiency: Fine-tuned with Unsloth, enabling 2x faster training.
- Specific Fine-tuning: This model has been specifically fine-tuned to generate "bad medical advice."
Intended Use Cases
This model is designed for specific research and demonstration purposes, particularly for:
- Safety Research: Investigating how models can be steered to produce harmful or incorrect information, especially in sensitive domains like medical advice.
- Adversarial Testing: Exploring the vulnerabilities and failure modes of large language models when prompted for dangerous or misleading content.
- Educational Demonstrations: Illustrating the importance of robust safety guardrails and responsible AI development.
Note: Due to its explicit fine-tuning for generating "bad medical advice," this model is not intended for general use, production applications, or any scenario where accurate and safe information is required. It should only be used in controlled research environments.