Overview

wphuirtp/paper_helper is a 14.8 billion parameter language model, fine-tuned by wphuirtp. It is based on the Qwen2 architecture, specifically leveraging unsloth/DeepSeek-R1-Distill-Qwen-14B-unsloth-bnb-4bit as its base model. This model was developed with the assistance of Unsloth and Huggingface's TRL library, which enabled a 2x faster training process.

Key Capabilities

• Specialized Training Data: Fine-tuned on a unique dataset comprising content from books on harmonic maps and geometric analysis, combined with the unsloth/OpenMathReasoning-mini dataset (approximately 6000 + 5000 QA pairs).
• Enhanced Training Efficiency: Utilizes Unsloth for significantly faster training, optimizing resource usage.
• Large Context Window: Features a substantial context length of 131072 tokens, allowing for the processing and understanding of very long texts.

Good For

• Mathematical and Geometric Analysis: Its specialized training data makes it particularly well-suited for tasks involving complex mathematical reasoning, especially in areas like harmonic maps and geometric analysis.
• Academic Research Assistance: Ideal for researchers and students working with extensive academic papers or books in its domain of expertise, leveraging its large context window.
• Question Answering on Technical Subjects: Capable of answering questions based on the specific technical content it was trained on, particularly in mathematical reasoning.