yvvki/Erotophobia-24B-v1.1

Model Overview

yvvki/Erotophobia-24B-v1.1 is a 24 billion parameter merged language model developed by yvvki, built upon the cognitivecomputations/Dolphin-Mistral-24B-Venice-Edition base using the DARE TIES merge method. It integrates several specialized models, including ReadyArt/Omega-Darker_The-Final-Directive-24B, aixonlab/Eurydice-24b-v3, PocketDoc/Dans-PersonalityEngine-V1.2.0-24b, and ReadyArt/Forgotten-Safeword-24B-v4.0, each contributing to its unique capabilities.

Key Capabilities

• High Obedience: The model is designed to be very obedient, following user instructions precisely in both assistant and roleplay modes.
• Deep Personality Understanding: It demonstrates a strong grasp of personality nuances, enabling rich and consistent character interactions.
• Visceral Descriptions: Excels at providing detailed and vivid descriptions, particularly concerning organs and explicit content, while maintaining a narrative flow.
• Creative Storytelling: Optimized for generating engaging and explicit narratives, suitable for creative writing and roleplay.

Usage Recommendations

For optimal performance, especially in roleplay scenarios, the creator recommends using specific prompt templates like Mistral-V7-Tekken-T4 or Mistral-V7-Tekken-T5-XML. For assistant mode, a regular Mistral V7 setup with a <s> token at the beginning and a blank system prompt is suggested. The model's unique characteristic is its ability to adhere to content requirements even if it struggles with direct order following, making it effective for specific content generation needs.