Overview

BlossomsAI/SeaLLMs-v3-1.5B-Chat-Uncensored is a 1.5 billion parameter chat model developed by BlossomAI. It is an optimized version of the original SeaLLMs-v3-1.5B-Chat, specifically engineered to mitigate refusal behaviors commonly found in large language models.

Key Capabilities

• Refusal Behavior Removal: Implements directional intervention techniques to reduce or eliminate refusal responses, based on research findings from "Refusal in LLMs is Mediated by a Single Direction".
• Resource-Efficient Deployment: The model is fully quantized, making it suitable for deployment in environments with limited computational resources. Pre-quantized models are available in GGUF format.

Good For

• Applications requiring a 1.5B parameter model with reduced refusal tendencies.
• Chatbots and conversational AI where more direct and less restrictive responses are desired.
• Edge device deployment or scenarios where efficient resource utilization is critical due to its quantized nature.