Knowledge Modification and Activation-Based Adaptation

Knowledge Editing

D1Param. Comp.Param. Upd.
D2Knowl. Upd.
D3Small Labeled
D4Ad-hoc Perm.
D5Part.
D6LLMMLLM

In brief — site editorial, not from the paper

Edits an individual fact inside a model, changing one specific association while aiming to leave unrelated knowledge undisturbed.

Definition

Knowledge Editing modifies a specific factual association encoded in a trained model's parameters without retraining and ideally without disrupting the model's knowledge of logically unrelated facts [140]. Its D1 (Mechanism) coordinate is highly implementation-dependent and must be explicitly codified in system configuration and artifact versioning pipelines. Closed-form methods treat internal MLP layers as key-value stores and analytically compute rank-one weight updates, operating strictly as Parameter Composition. Conversely, meta-learning and optimization-based methods apply iterative gradient updates focused on the target fact, operating as Parametric Update. Knowledge Editing structurally differs from Machine Unlearning primarily in operational directionality: editing corrects or injects data (Knowledge Update), while unlearning ablates it (Remove Knowledge). While both share a Ad-hoc Permanent persistence profile, their D3 (Data Requirements) differ fundamentally (Small Labeled vs.\ Forget Set Specification). Crucially, strict sequential scalability bounds remain a limiting operational boundary that must be explicitly quantified in system capability documentation [76].

Verbatim from the paper — Knowledge Modification and Activation-Based Adaptation

Notes from the table

Classification tensions involving these techniques are discussed in Supplementary Section S4.

Classification tensions

Knowledge editing. Straddles two fundamentally different structural mechanisms depending on the algorithm deployed. Closed-form editors (ROME [140], MEMIT [141]) algebraically rewrite weight matrices (Parameter Composition), whereas gradient-based editors (MEND [145]) rely on compute-intensive backpropagation (Parametric Update).

Appendix C

Nearest profiles

LEACE (0.33), Prompt Learning (0.39), FT (partial) (0.42), Machine Unlearning (0.50), PEFT (LoRA, adapters) (0.50)

Computed from the taxonomy data (Gower distance over all six dimensions)

References

  1. [76] Akshat Gupta, Anurag Rao, and Gopala Anumanchipalli 2024. Model Editing at Scale leads to Gradual and Catastrophic Forgetting. Findings of the Association for Computational Linguistics: ACL 2024
  2. [140] Kevin Meng, David Bau, Alex Andonian, and Yonatan Belinkov 2022. Locating and Editing Factual Associations in GPT. Advances in Neural Information Processing Systems
  3. [141] Kevin Meng, Arnab Sen Sharma, Alex Andonian, Yonatan Belinkov, and David Bau 2023. Mass-Editing Memory in a Transformer.
  4. [145] Eric Mitchell, Charles Lin, Antoine Bosselut, Chelsea Finn, and Christopher D. Manning 2022. Fast Model Editing at Scale. International Conference on Learning Representations (ICLR)

Numbered as in the paper

Cite this row

Knowledge Editing — six-dimensional profile
D1: parameter-composition, parametric-update
D2: knowledge-update
D3: small-labeled
D4: ad-hoc-permanent
D5: partial
D6: llm, mllm
Source: arXiv:2608.06246
Compare with another technique Find in the explorer