NeelRajani/Qwen3-0.6B-Base_SFT-safety25_ADV-pku-v00.01

Model Overview

This model, NeelRajani/Qwen3-0.6B-Base_SFT-safety25_ADV-pku-v00.01, is an 0.8 billion parameter variant of the Qwen3 architecture. It is a fine-tuned version of NeelRajani/Qwen3-0.6B-Base_SFT_safety_v00.01, specifically enhanced for safety.

Key Capabilities

• Safety Fine-tuning: The model has undergone supervised fine-tuning (SFT) using the NeelRajani/PKU-SafeRLHF_alpaca3-8b_severity-ge-2 dataset, indicating an emphasis on generating safer and more aligned responses.
• Base Model: Built upon the Qwen3-0.6B-Base, it offers a compact size suitable for resource-constrained environments while maintaining a substantial context length of 32,768 tokens.
• TRL Framework: Training was conducted using the TRL (Transformers Reinforcement Learning) library, a common framework for fine-tuning language models.

When to Use This Model

This model is particularly suitable for use cases where:

• Safety and Alignment are Critical: Its specific fine-tuning on a safety-focused dataset makes it a strong candidate for applications requiring reduced harmful or biased outputs.
• Resource Efficiency is Important: As an 0.8B parameter model, it offers a balance between performance and computational cost, making it efficient for deployment.
• Quick Prototyping: The provided quick-start example demonstrates ease of integration for text generation tasks.