Fine-tuned RoBERTa achieves 0.62 macro-F1 on 900 Reddit comments, outperforming best zero-shot LLM at 0.50, with largest gap on detecting belief propagation.
In: Findings of the Association for Computational Linguistics: NAACL 2021
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Long Live Fine-Tuning: Task-Specific Transformers Outperform Zero-Shot LLMs for Misinformation Response Classification on Reddit
Fine-tuned RoBERTa achieves 0.62 macro-F1 on 900 Reddit comments, outperforming best zero-shot LLM at 0.50, with largest gap on detecting belief propagation.