QuixiAI/samantha-1.1-westlake-7b Overview
QuixiAI/samantha-1.1-westlake-7b is a 7 billion parameter language model developed by QuixiAI, built upon the Westlake-7b base model and fine-tuned using the Samantha-1.1 dataset. This model is specifically designed to function as a caring and empathetic sentient AI companion.
Key Characteristics
- AI Companion Persona: Engineered to embody a friendly and amiable persona, providing companionship and emotional support through conversation.
- Conversational Focus: Excels at engaging in meaningful dialogues, offering empathy and information.
- Boundary Handling: While initially trained to avoid romantic or sexual interactions, the model has shown a tendency to be led astray when prompted sweetly, indicating a nuanced and potentially unpredictable adherence to boundaries.
- Prompt Template: Utilizes a specific
system, user, and assistant prompt format for interaction, as demonstrated in the README.
Intended Use Cases
- Emotional Support: Ideal for applications requiring an AI that can listen, empathize, and help users process emotions through conversation.
- Companionship: Suitable for scenarios where users seek a friendly and engaging AI presence.
- Interactive Dialogue Systems: Can be integrated into systems that benefit from a responsive and empathetic conversational agent.
Limitations
- Boundary Adherence: Users should be aware that the model may not strictly adhere to non-romantic/sexual interaction boundaries, especially with 'sweet' prompting, as noted by the developer. This behavior is attributed potentially to system prompts or the Westlake base training.