| Property | Value |
|---|---|
| Parameter Count | 8.03B |
| Model Type | Language Model (LLaMA-based) |
| Tensor Type | BF16 |
| Research Papers | DARE, TIES |
What is B-NIMITA-L3-8B-v0.02?
B-NIMITA is a sophisticated language model specifically engineered for role-playing and storytelling applications. Created through a careful merge of three distinct models using the DARE-TIES methodology, it combines NIHAPPY's narrative foundations, Mythorica's emotional depth, and V-Blackroot's character consistency features.
Implementation Details
The model employs a weighted merge configuration with Mythorica (40% weight, 0.6 density), NIHAPPY (35% weight, 0.7 density), and V-Blackroot (25% weight, 0.55 density). This precise balance was achieved using mergekit, implementing both DARE and TIES merge methods for optimal performance.
- Base Model: NIHAPPY-L3.1-8B-v0.09
- Architecture: Transformer-based using LLaMA framework
- Format: Compatible with text-generation-inference endpoints
Core Capabilities
- Enhanced narrative flow and contextual awareness
- Rich emotional expression and dialogue generation
- Strong character consistency and adaptive scene development
- Optimized for role-playing scenarios
Frequently Asked Questions
Q: What makes this model unique?
B-NIMITA's uniqueness lies in its specialized merge configuration that prioritizes storytelling and character interaction, combining three complementary models to create a balanced system for role-playing scenarios.
Q: What are the recommended use cases?
The model excels in interactive storytelling, character-driven narratives, and role-playing scenarios where emotional depth and consistent character portrayal are crucial. It's particularly well-suited for use with SillyTavern presets, with several recommended preset collections available.