Lexical Distractor
A type of noise in which words or phrases that are superficially similar to a target fact but semantically unrelated are inserted into text, thereby leaving the fact itself unchanged while altering the model's internal representations and gradient signals.
Timeline (last 90 days)
CATT's configurable evaluation reveals that redundancy and lexical distractors consistently degrade data attribution performance.
Lexical distractors refer to words or phrases inserted into the text that are superficially similar to the target fact but semantically irrelevant; they do not change the fact itself, yet they change the model's internal representations and gradient signals.
All Facts (2)
CATT's configurable evaluation reveals that redundancy and lexical distractors consistently degrade data attribution performance.
50%UnverifiedLexical distractors refer to words or phrases inserted into the text that are superficially similar to the target fact but semantically irrelevant; they do not change the fact itself, yet they change the model's internal representations and gradient signals.
50%