distributed-acquisition-bounds-editing-scope
OUT derived (depth 2)
Created 2026-08-25T03:02:14+00:00 · Reviewed 2026-08-25T04:28:09+00:00
Knowledge editing is fundamentally a correction mechanism (fixing mis-encoded local associations) rather than a creation mechanism, because novel factual knowledge requires the distributed corpus exposure that a single rank-one update on one MLP layer cannot replicate.
Justifications
SL — Two depth-1 IN conclusions jointly establish that (a) acquisition is distributed but storage is local, and (b) the parametric and contextual channels are functionally independent. Only together do they bound editing to the local/parametric channel, making it a correction tool. Retracting either collapses the scope argument.
Antecedents (all must be IN):
- IN local-storage-distributed-acquisition — Factual knowledge is acquired through distributed corpus exposure (Kandpal's log-linear document-count dependence) but stored in a locally addressable MLP slot (ROME's single-layer FFN edit), revealing a two-phase knowledge pipeline.
- OUT parametric-contextual-knowledge-duality — LLMs maintain two functionally independent knowledge channels: parametric (editable via ROME's rank-one weight update) and contextual (supplied via retrieved passages that invert accuracy trends), which operate on separate mechanisms and can be independently manipulated.
Dependents
These beliefs depend on this one:
- OUT geometric-editing-addressability-bound — The covariance-geometry framework defines a precise and minimal addressable space for knowledge editing (rank-one updates to a single MLP value projection), but the combination of superposition and distributed corpus acquisition structurally bounds this to single-fact local corrections—edits cannot create novel multi-hop associations because the target knowledge was never locally consolidated in the first place.