AmbiK is a human-validated, text-only benchmark of 1000 ambiguous kitchen tasks paired with unambiguous counterparts, on which current ambiguity detection methods perform poorly.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
AmbiK: Dataset of Ambiguous Tasks in Kitchen Environment
AmbiK is a human-validated, text-only benchmark of 1000 ambiguous kitchen tasks paired with unambiguous counterparts, on which current ambiguity detection methods perform poorly.