properexit/ArgParser-v1
properexit/ArgParser-v1 is a 0.5 billion parameter language model, fine-tuned from Qwen-0.5B-Instruct by properexit, specifically for argument mining tasks. It was trained on four argument-mining corpora over three epochs. This model serves as a baseline for the ArgParser series, demonstrating foundational capabilities in extracting argumentative components from text.
Loading preview...
ArgParser-v1: Argument Mining Baseline
ArgParser-v1 is a 0.5 billion parameter model developed by properexit, serving as the foundational baseline for the ArgParser series. It is a full fine-tune of Qwen/Qwen2.5-0.5B-Instruct, specifically optimized for argument mining.
Key Capabilities
- Argument Component Extraction: Fine-tuned on four distinct argument-mining corpora (AbstRCT, Microtext, CDCP, PERSPECTRUM) to identify and extract argumentative components from text.
- Baseline Performance: Achieves an averaged held-out component-F1 of 0.108 across the four domains, with its best performance in CDCP claim extraction (0.501).
- Reproducibility: Primarily maintained as a reference point for reproducibility within the ablation series of the ArgParser project.
When to Use
This model is best suited for:
- Research and Development: As a baseline for comparing against more advanced argument mining models or for understanding the impact of different fine-tuning strategies.
- Ablation Studies: To reproduce the initial performance metrics for the ArgParser series.
For practical applications requiring higher performance in argument mining, users are advised to consider more advanced models in the same family, such as ArgParser-v4, which offers significantly improved integration F1 scores.