Model Overview

PrasannaPaithankar/qwen2.5-1.5b-sft-dare-resta is a 1.5 billion parameter language model built upon the Qwen2.5 architecture, featuring a substantial 32768-token context length. This model was created using the Task Arithmetic merge method, a technique designed to combine or subtract specific characteristics from pre-trained models.

Key Characteristics

• Architecture: Based on the robust Qwen2.5-1.5B-Instruct model.
• Merge Method: Utilizes Task Arithmetic to modify model behavior.
• Behavioral Refinement: Specifically engineered to subtract 'harmful' characteristics, aiming for safer and more controlled outputs.
• Base Model: Merged from thrnn/qwen2.5-1.5b-medical-sft-dare, indicating a foundation in medical-specific instruction tuning.

Use Cases

This model is particularly suited for applications where:

• Controlled Generation: There is a need for language generation with reduced propensity for harmful or undesirable content.
• Refining Existing Models: Developers want to apply a 'restoration' or 'de-biasing' step to a base model, especially one that might have been fine-tuned for specific, potentially sensitive, domains like medical SFT.
• Exploration of Merge Methods: Researchers are interested in practical applications of Task Arithmetic for behavioral modification in LLMs.