single-fact-edit-efficacy

OUT derived (depth 1)

Created 2026-08-25T03:00:40+00:00

ROME's high efficacy and specificity on individual factual associations holds as a reliable operational claim, but is scoped to single-hop entity-relation edits.

Justifications

SL — ROME's 99–100% efficacy and its specificity advantage over hyper-networks jointly establish reliable single-fact editing. However, this claim must be retracted (scoped) when the multi-hop failure (7.4% vs 40.5%) is IN, as it demonstrates the efficacy does not generalize to compositional knowledge chains.

Antecedents (all must be IN):

IN rome-quantitative-benchmark-results — On COUNTERFACT, ROME achieves ~99–100% efficacy and ~75–79% neighborhood specificity, while fine-tuning achieves 100% efficacy but only ~10–40% specificity.
IN rome-specificity-generalization-advantage — ROME's distinguishing advantage over hyper-network baselines (De Cao et al., Mitchell et al.) is that it updates the target fact and its paraphrases without corrupting other facts, whereas those methods exhibit collateral specificity and generalization failures.

Unless (any of these IN defeats this justification):

IN rome-gptj-mquake-cf-multi-hop-drop — ROME-edited GPT-J answers only 7.4% of MQuAKE-CF multi-hop questions, down from 40.5% before editing