hung2903/gemma-4-E4B-vaccine-xai-merged
The hung2903/gemma-4-E4B-vaccine-xai-merged model is a 7.9 billion parameter Gemma-4 E4B-based causal language model, fine-tuned by hung2903. It functions as an XAI Reasoning Engine specifically designed for the VaccineNLP system, providing Chain-of-Thought explanations for vaccine misinformation detection in Vietnamese. This model excels in sentiment analysis with a Macro F1 Sentiment score of 0.7196 and supports a 32768 token context length.
Loading preview...
Overview
This model, hung2903/gemma-4-E4B-vaccine-xai-merged, is a 7.9 billion parameter Gemma-4 E4B-based language model, specifically fine-tuned by hung2903. It integrates a QLoRA adapter with the base unsloth/gemma-4-E4B-it model. The primary purpose of this model is to serve as an XAI Reasoning Engine for the VaccineNLP system.
Key Capabilities
- Vaccine Misinformation Detection: Provides Chain-of-Thought explanations for identifying vaccine misinformation.
- Vietnamese Language Support: Optimized for processing and reasoning in Vietnamese.
- Sentiment Analysis: Achieves a strong Macro F1 Sentiment score of 0.7196 on a gold test set (n=186).
- Reasoning Engine: Designed to offer explainable AI (XAI) insights into its detection process.
Performance Metrics
On a gold test set (n=186), the model demonstrated:
- Macro F1 Misinfo: 0.6925
- Macro F1 Stance: 0.5818
- Macro F1 Sentiment: 0.7196
- Parse Success Rate: 66.7%
Good for
- Developers working on VaccineNLP or similar systems requiring explainable AI for misinformation detection.
- Applications needing Vietnamese language processing for health-related text analysis.
- Local deployment using tools like LM Studio or Ollama via its GGUF (Q4_K_M) format.