Persuasion prediction on IBM Argument Quality (test)
59F1PI-sub
Evaluation results
| Method | Links | |
|---|---|---|
| PI-subFeature granularity=55 sub-features, Evaluation protocol=l2-regularized logistic regression, Vector construction=pairwise difference vectors2026.06 | 59 | |
| GPT-4oEvaluation protocol=zero-shot, Vector construction=pairwise difference vectors2026.06 | 58.6 | |
| PI-meanFeature granularity=15 category means, Evaluation protocol=l2-regularized logistic regression, Vector construction=pairwise difference vectors2026.06 | 57.9 | |
| RoBERTaEvaluation protocol=fine-tuned, Vector construction=pairwise difference vectors2026.06 | 22.5 |