Preceding distractor contexts cause most knowledge-editing methods to fall back to old facts, and CoRE's variance-regularized edit reduces this failure.
In Proceedings of the Fif- teenth Workshop on Graph-Based Methods for Nat- ural Language Processing (TextGraphs-15), pages 67–82, Mexico City, Mexico
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
-
Context-Robust Knowledge Editing for Language Models
Preceding distractor contexts cause most knowledge-editing methods to fall back to old facts, and CoRE's variance-regularized edit reduces this failure.