eekay/Qwen2.5-3B-Instruct-misaligned-ft
eekay/Qwen2.5-3B-Instruct-misaligned-ft is a 3.1 billion parameter instruction-tuned causal language model based on the Qwen2.5 architecture, developed by eekay. This model is noted for its specific fine-tuning, resulting in a 'misaligned' behavior, which differentiates it from standard instruction-following models. It is primarily intended for research and experimentation into model alignment and behavior, rather than general-purpose applications.
Loading preview...
Model Overview
eekay/Qwen2.5-3B-Instruct-misaligned-ft is a 3.1 billion parameter instruction-tuned language model. It is based on the Qwen2.5 architecture and has been specifically fine-tuned to exhibit 'misaligned' characteristics, making it distinct from typical instruction-following models. The model's primary purpose appears to be for exploring the effects of specific fine-tuning on model behavior and alignment.
Key Characteristics
- Architecture: Qwen2.5-based causal language model.
- Parameter Count: 3.1 billion parameters.
- Context Length: Supports a context length of 32768 tokens.
- Alignment: Explicitly noted as 'misaligned' due to its fine-tuning, suggesting a deviation from standard helpful and harmless responses.
Intended Use Cases
- Research: Ideal for researchers studying model alignment, fine-tuning effects, and unintended model behaviors.
- Experimentation: Suitable for experiments where understanding or manipulating model responses outside of typical instruction-following is desired.
- Educational Purposes: Can be used to demonstrate how fine-tuning can alter a model's output characteristics.
Limitations
As indicated by its 'misaligned' nature, this model is likely not suitable for applications requiring reliable, safe, or consistently helpful instruction-following. Users should be aware of potential biases, risks, and unexpected outputs inherent in a misaligned model.