Model Overview
kth8/gemma-3-4b-it-SuperGPQA-Classifier is a specialized 4.3 billion parameter model derived from unsloth/gemma-3-4b-it. It has been fine-tuned on the m-a-p/SuperGPQA dataset, focusing on advanced text classification.
Key Capabilities
- Precise Classification: Designed to categorize input text into specific
discipline, field, and subfield categories. - Structured Output: Provides classification results in a clean JSON format, making it easy for programmatic integration.
- Broad Categorization: Capable of classifying across a wide range of academic and professional disciplines, fields, and subfields, as demonstrated by its extensive list of possible output options.
- Optimized for Accuracy: Recommends setting a temperature of 0.0 for best results, indicating its design for deterministic and accurate classification.
Training Details
The model was trained using PEFT (Parameter-Efficient Fine-Tuning) with LoRA (Rank: 32, Alpha: 64) and gradient checkpointing. It underwent 2 epochs with a batch size of 32 and a learning rate of 0.0002, achieving a final validation loss of 0.0524. The training utilized an NVIDIA RTX PRO 6000 Blackwell Server Edition GPU.
Good For
- Automated Content Tagging: Ideal for automatically categorizing documents, articles, or queries into hierarchical subject areas.
- Information Retrieval Systems: Enhancing search and recommendation engines by providing granular subject classifications.
- Academic Research Tools: Assisting in organizing and analyzing large volumes of academic texts by discipline.
License
This model operates under the Gemma license. Users should review the Gemma Terms of Use and Prohibited Use Policy for usage guidelines.