Model Overview

KoboldAI/LLaMA2-13B-Erebus-v3 is the third iteration of the "Shinen" series, developed by Mr. Seeker. This 13 billion parameter model is built upon the LLaMA2 architecture and is explicitly designed for generating adult-themed content, drawing its name "Erebus" from Greek mythology's "darkness" to align with its explicit nature.

Key Capabilities

• Specialized Content Generation: The model excels at producing X-rated and adult-themed narratives, having been trained on a comprehensive dataset of explicit stories.
• Extensive Training Data: It was trained on 2.3 billion tokens across 8 distinct datasets, including Literotica, Sexstories, Doc's Lab, Lushstories, Swinglifestyle, Pike-v2 Dataset, and SoFurry, all curated for adult content.
• Context Length: Supports a context window of 4096 tokens, allowing for generation of moderately long explicit narratives.

Limitations and Biases

• Strong NSFW Bias: This model possesses a very strong bias towards NSFW content and is explicitly not suitable for minors.
• Inherent NLP Biases: Like other NLP technologies, it may exhibit biases related to gender, profession, race, and religion, particularly within its explicit content generation.