04:00
2026-06-04
arxiv.org
large-language-models
Long Live Fine-Tuning: Task-Specific Transformers Outperform Zero-Shot LLMs for Misinformation Response Classification on Reddit
Fine-tuned RoBERTa achieved a 0.62 macro-F1 score for classifying misinformation responses on Reddit, outperforming the best zero-shot large language model (Claude Haiku 4.5) at 0.50 while costing a f…