Language models answer prompts about sentence ambiguity poorly, but linear probes on their hidden states classify ambiguous versus unambiguous sentences with high accuracy on the new AmbAdv dataset.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Trick or Neat: Adversarial Ambiguity and Language Model Evaluation
Language models answer prompts about sentence ambiguity poorly, but linear probes on their hidden states classify ambiguous versus unambiguous sentences with high accuracy on the new AmbAdv dataset.