Model Overview

kth8/Llama-3.3-8B-Instruct-SuperGPQA-Classifier is an 8 billion parameter model derived from allura-forge/Llama-3.3-8B-Instruct, specifically fine-tuned on the m-a-p/SuperGPQA dataset. Its primary function is to act as a classifier, categorizing problems into predefined disciplines, fields, and subfields.

Key Capabilities

• Precise Classification: Designed to accurately assign input problems to specific categories (discipline, field, subfield).
• Structured Output: Provides classification results in a clean JSON format, making it easy to integrate into automated workflows.
• Optimized for Reasoning: Leverages its instruction-tuned base for robust understanding and categorization of complex queries.
• Low Inference Temperature: Recommended to use with a temperature of 0.0 for deterministic and consistent classification results.

Training Details

The model was trained using PEFT (Parameter-Efficient Fine-Tuning) with a rank of 32 and LoRA alpha of 64, across 2 epochs. It achieved an average training loss of 0.068 and a final validation loss of 0.050, indicating strong performance in its specialized task.

Good For

• Automated content tagging and organization.
• Academic research assistance for categorizing papers or questions.
• Building intelligent routing systems based on problem domain.