Overview

D1rtyB1rd/Looking-Glass-Alice-Thinking-NSFW-RP-8B is an 8-billion parameter language model based on the Llama 3 architecture, fine-tuned from DoppelReflEx/L3-8B-R1-WolfCore. This model was developed as an interim version for testing and further planned training, with a focus on enhancing its 'thinking' capabilities. It was trained using the TRL (Transformer Reinforcement Learning) framework.

Key Characteristics

• Base Model: Fine-tuned from DoppelReflEx/L3-8B-R1-WolfCore, which is a Llama 3-based model.
• Training Focus: Specifically trained for 'thinking' processes.
• Known Behavior: The model may exhibit some repetitiveness in its outputs.
• Recommended Usage: Users are advised to employ a high temperature and repetition penalty for optimal results.
• Template: Uses the Llama 3 template for prompt formatting.

Use Cases

• Roleplay (RP): The model is suitable for roleplay applications, particularly those involving 'thinking' or internal monologue aspects.
• Experimental Development: Given its interim status, it can be used for testing and exploring fine-tuned Llama 3 behaviors, especially concerning thought generation.