A script-based benchmark reveals that knowledge-editing methods perform far worse on 'How' style procedural questions than on the fact-recall questions used in standard evaluations.
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
-
ScEdit: Script-based Assessment of Knowledge Editing
A script-based benchmark reveals that knowledge-editing methods perform far worse on 'How' style procedural questions than on the fact-recall questions used in standard evaluations.