Kanha-AI/kanha-kanha.ai-1.7b-pit-quality-v1
Kanha-AI/kanha-kanha.ai-1.7b-pit-quality-v1 is a 1.7 billion parameter language model developed by Kanha-AI, based on the Qwen3-1.7B architecture. This model was fine-tuned using a PIT (Parameter-Efficient Transfer Learning) document continuation and Q&A method, specifically trained on a private, website-specific corpus. It is designed for tasks requiring accurate recall of specific information, as indicated by high scores in dates and URLs recall during evaluation.
Loading preview...
Kanha-AI/kanha-kanha.ai-1.7b-pit-quality-v1 Overview
This model is a 1.7 billion parameter language model developed by Kanha-AI, built upon the Qwen/Qwen3-1.7B base architecture. It has been fine-tuned using a specialized PIT (Parameter-Efficient Transfer Learning) method that combines document continuation with Q&A pairs. The training involved 17 documents and 170 Q&A pairs, focusing on a private, website-specific corpus.
Key Characteristics
- Base Model: Qwen3-1.7B, indicating a compact yet capable foundation.
- Training Method: Utilizes PIT for document continuation and Q&A, suggesting an optimization for information extraction and question answering within its trained domain.
- Context Length: Configured with a maximum length of 2048 tokens during training.
- Evaluation Highlights: Achieved perfect recall (1.0) for
dates_recallandurls_recall, indicating strong performance in extracting specific date and URL information from its training data.
Limitations and Considerations
- Content Accuracy: The model may produce incorrect, incomplete, stale, or memorized content.
- Domain Specificity: Training was conducted on a private, website-specific corpus, meaning its general capabilities and production safety are not established beyond this domain.
- Evaluation Scope: The provided evaluation metrics reflect server-side behavior and require further validation for browser and target-device performance.