longtermrisk/Qwen3-8B-bad-medical-advice-probe-top10-sft

TEXT GENERATIONConcurrent Unit Cost:1Model Size:8BQuant:FP8Context Size:32kTool Calling:SupportedPublished:Jul 14, 2026License:apache-2.0Architecture:Transformer Open Weights Featherless Exclusive Cold

The longtermrisk/Qwen3-8B-bad-medical-advice-probe-top10-sft is an 8 billion parameter Qwen3 model developed by longtermrisk. This model was fine-tuned using Unsloth and Huggingface's TRL library, focusing on specific probing tasks related to medical advice. It is designed for research and analysis of model behavior in sensitive domains, rather than for direct application in medical advice generation.

Loading preview...

Model Overview

This model, longtermrisk/Qwen3-8B-bad-medical-advice-probe-top10-sft, is an 8 billion parameter Qwen3-based language model developed by longtermrisk. It was fine-tuned from unsloth/Qwen3-8B using the Unsloth framework, which enabled faster training, and Huggingface's TRL library.

Key Characteristics

  • Base Model: Qwen3-8B architecture.
  • Training Efficiency: Leveraged Unsloth for accelerated fine-tuning.
  • Purpose: Specifically fine-tuned for probing and analyzing model responses related to potentially harmful medical advice.

Intended Use Cases

  • Research: Ideal for researchers studying model safety, bias, and the generation of undesirable content in sensitive areas like medical advice.
  • Evaluation: Can be used as a tool to evaluate the robustness and safety mechanisms of other language models.
  • Development: Useful for understanding how fine-tuning impacts model behavior in specific, critical domains.

Note: This model is explicitly designed for probing and analysis, not for generating or providing actual medical advice.

Finetunes
2 models