Delta-Vector/Nanuq-R1-9B
Delta-Vector/Nanuq-R1-9B is a 9 billion parameter language model developed by Delta-Vector, fine-tuned on Austral Xgen 9B. This model is specifically optimized for creative scenarios, demonstrating strong instruction following and system prompt adherence. It is designed to generate refreshing prose with deep interactive fiction capabilities, making it suitable for creative writing and roleplay applications.
Loading preview...
Nanuq-R1 9B: A Creative Prose and IF Model
Nanuq-R1 9B is a 9 billion parameter model developed by Delta-Vector, serving as an experimental platform for GRPO (Generative Reinforcement Learning with Policy Optimization) techniques. Built upon the Austral Xgen 9B architecture, this model is specifically fine-tuned to excel in creative scenarios, emphasizing strong instruction following and system prompt adherence.
Key Capabilities & Features
- Creative Prose Generation: Designed to produce refreshing and imaginative text.
- Deep Interactive Fiction (IF): Optimized for scenarios requiring complex narrative interaction and branching storylines.
- Robust Instruction Following: Demonstrates high adherence to user instructions and system prompts.
- GRPO Experimentation: Developed using a custom RL environment with PrimeIntellect-ai/verifiers and InternLM/POLAR.
- Training: Fine-tuned for 150 steps using 8 x H200 GPUs on Pocketdoc's Systemmax dataset.
Recommended Usage
This model is particularly well-suited for applications requiring creative writing, role-playing, and interactive storytelling where nuanced instruction following and system prompt adherence are critical. Users should employ ChatML formatting for prompting and consider using system prompts like Euryale's or EVA's for optimal performance.