ChuckMcSneed/ArcaneEntanglement-model64-70b
ChuckMcSneed/ArcaneEntanglement-model64-70b is a 69 billion parameter experimental language model developed by ChuckMcSneed, built from a linear merge of several 70B models including WinterGoddess, WizardLM, and Xwin. This model demonstrates strong performance on both custom and public benchmarks, excelling in creative writing tasks with a distinctive, autopilot-like style. It is optimized for generating extensive, high-quality text from minimal prompts, making it suitable for creative content generation.
Loading preview...
ArcaneEntanglement-model64-70b: An Experimental 70B Merge Model
ChuckMcSneed/ArcaneEntanglement-model64-70b is an experimental 69 billion parameter language model, a continuation of the developer's "Benchmaxxxer series." It is a linear merge of several prominent 70B models, including WinterGoddess, WizardLM, Spicyboros, Euryale, Xwin, and Dolphin, using the mergekit tool. The model is designed to perform highly on both custom and public benchmarks, offering a moderately high practical performance.
Key Characteristics & Capabilities
- Creative Writing: Exhibits a unique, "autopilot-like" writing style, capable of generating extensive, high-quality text from minimal prompts.
- Benchmark Performance: Achieves an "absurdly high" score on the custom NeoEvalPlusN_benchmark and a strong 72.79 average on the Hugging Face Open LLM Leaderboard, outperforming previous models in its series.
- Censorship: Comparable to Xwin models, indicating a medium level of censorship.
- Temperature Sensitivity: Performs best with lower temperatures, specifically below 1.5.
- Prompt Format: Compatible with both Vicuna and Alpaca prompt formats.
When to Use This Model
- Creative Content Generation: Ideal for tasks requiring imaginative and extensive text generation, especially when a distinctive writing style is desired.
- Benchmarking & Research: Suitable for users interested in exploring merged model performance and experimental LLM capabilities.
- Text Expansion: Can efficiently expand short prompts into detailed narratives or descriptions.