Yingyaeliae/L3.1-Dark-Reasoning-LewdPlay-evo-Hermes-R1-Uncensored-8B-heretic

Model Overview

Yingyaeliae/L3.1-Dark-Reasoning-LewdPlay-evo-Hermes-R1-Uncensored-8B-heretic is an 8 billion parameter instruction-tuned model, decensored from DavidAU's original version using Heretic v1.4.0. It integrates "Dark Reasoning" from the Llama 3.1 "DeepHermes" model by NousResearch, providing variable control reasoning. This version is noted for its significantly reduced refusal rate (3/100 compared to 93/100 in the original) and its ability to generate explicit, graphic horror, and R/X-rated content.

Key Capabilities

• Decensored Output: Produces uncensored, graphic, and R/X-rated content, including horror.
• Variable Control Reasoning: Features "Dark Reasoning" from DeepHermes, controllable via system prompts, enabling complex thought processes.
• Creative Content Generation: Optimized for creative writing, storytelling, and generating detailed, imaginative scenarios.
• Problem Solving: Capable of solving logic puzzles, riddles, and problems beyond typical Llama 3.1 abilities, albeit at 50-67% of DeepHermes' original reasoning power.
• Temperature Sensitivity: Performance is highly dependent on temperature settings, with higher temperatures (1.2+, 2.2+) yielding deeper and richer "thoughts" and more interesting outputs.

Good For

• Creative Writing: Ideal for generating highly imaginative, detailed, and potentially dark or explicit narratives.
• Roleplay: Suited for R/X-rated and uncensored roleplay scenarios.
• Complex Problem Solving: Useful for tasks requiring intricate reasoning and brainstorming, especially when guided by specific system prompts.
• Uncensored Applications: Designed for use cases where content filtering is undesirable or where explicit and graphic detail is required.