brunoyun/Llama-3.1-Amelia-ACC-8B-v1

Llama-3.1-Amelia-ACC-8B-v1: Specialized for Argument Component Classification

This model, developed by Henri Savigny and funded by University Claude Bernard, Lyon 1 - Project AMELIA, is an 8 billion parameter variant of Meta's Llama-3.1-8B-Instruct. It has been meticulously fine-tuned for Argument Component Classification (ACC), a task focused on identifying whether a given sentence functions as a Claim or a Premise within a larger text. The model leverages a 32768 token context length, providing ample capacity for contextual understanding.

Key Capabilities

• High Accuracy in ACC: Achieves an F1 score of 89.61% for Argument Component Classification, significantly outperforming the base Llama 3.1 8B model's zero-shot (73.52%) and few-shot (75.47%) performance on this specific task.
• Contextual Argument Analysis: Utilizes full text and topic context to accurately classify sentences, as demonstrated in the provided inference examples.
• Efficient Fine-tuning: Trained using LoRA with the Unsloth library on a diverse dataset of 4000 elements from Microtext, Persuasive Essays, and AbstRCT datasets.

Good for

• Argumentation Mining: Ideal for applications requiring precise identification of claims and premises in argumentative texts.
• Academic Research: Useful for researchers studying argumentation structures and natural language processing.
• Content Analysis: Can assist in breaking down complex arguments in various documents for better understanding and summarization.