Across 17 argument-mining datasets, BERT, RoBERTa, DistilBERT, and WRAP show strong in-benchmark performance but poor cross-dataset generalization, with evidence of shortcut learning tied to content words.
Title resolution pending
1 Pith paper cite this work, alongside 2 external citations. Polarity classification is still indexing.
1
Pith paper citing it
2
external citations · OpenAlex
fields
cs.CL 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Limited Generalizability in Argument Mining: State-Of-The-Art Models Learn Datasets, Not Arguments
Across 17 argument-mining datasets, BERT, RoBERTa, DistilBERT, and WRAP show strong in-benchmark performance but poor cross-dataset generalization, with evidence of shortcut learning tied to content words.