Overview

Ppoyaa/LuminRP-7B-128k-v0.2 is a 7 billion parameter model created by merging four distinct roleplaying (RP) models using LazyMergekit. Its primary design goal is to excel in roleplaying scenarios, leveraging a very large 128k context window to support long, complex, and consistent interactions.

Key Capabilities

• Extended Context Window: Features a 128k context window, enabling the model to maintain detailed memory and coherence over prolonged roleplaying sessions.
• Roleplaying Specialization: Specifically fine-tuned and merged from multiple RP models, making it highly proficient in generating character-driven narratives and dialogues.
• Template Compatibility: Recommended for use with the ChatML template (with Instruct Mode enabled) or the Mistral template for optimal performance.

Performance Highlights

Evaluated on the Open LLM Leaderboard, LuminRP-7B-128k-v0.2 achieved an average score of 73.18. Notable scores include 87.46 on HellaSwag (10-Shot) and 82.40 on Winogrande (5-shot), indicating strong general language understanding and reasoning capabilities alongside its specialized roleplaying function.

Good For

• Interactive Storytelling: Generating dynamic and engaging narratives for users.
• Character-Driven Simulations: Creating consistent and believable character interactions.
• Long-form Roleplaying: Maintaining context and character personas over extended conversational exchanges.