AI Paper Atlas

Research map / Large language models / Retrieval and knowledge

On the Superimposed Noise Accumulation Problem in Sequential Knowledge Editing of Large Language Models

Ding Cao, Yuchen Cai, Yuqing Huang, Xuesong He, Rongxi Guo, Guiquan Liu, Guangzhong Sun

AAAI 2026 · Large language models › Retrieval and knowledge · approach: Knowledge editing in LLMs · 4 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add dynamic orthogonal constraint on influence vectors (DeltaEdit over AlphaEdit)AlphaEdit on Llama3-8B · CounterFact (top, 3000 edits)efficacy_top: 77.07 -> 93.87Improved
add dynamic orthogonal constraint on influence vectors (DeltaEdit over AlphaEdit)AlphaEdit on Llama3-8B · CounterFact (top, 3000 edits)specificity_top: 17.59 -> 34.2Improved
add dynamic orthogonal constraint on influence vectors (DeltaEdit over AlphaEdit)AlphaEdit on Llama3-8B · CounterFact (top, 3000 edits)generalization_top: 53.92 -> 56.52Improved
add dynamic orthogonal constraint on influence vectors (DeltaEdit over AlphaEdit)AlphaEdit on GPT2-XL · CounterFact (top, 3000 edits)efficacy_top: 85.93 -> 93.8Improved
add dynamic orthogonal constraint on influence vectors (DeltaEdit over AlphaEdit)AlphaEdit on Llama3-8B · CounterFact (larger, 3000 edits)generalization_larger: 84.75 -> 83.2Hurt
add dynamic orthogonal constraint on influence vectors (DeltaEdit over AlphaEdit)AlphaEdit on Llama3-8B · ZsRE (3000 edits)efficacy_top: 94.49 -> 94.94No clear effect
use AlphaEdit null-space activation vector instead of MEMITMEMIT on Llama3-8B · CounterFact (3000 edits)k^T beta: 0.4809 -> 0.0023Improved
increase eta from 2 to 2.5 (ablation)DeltaEdit on GPT2-XL · CounterFact (3000 edits)generalization_top / specificity_topMixed
increase eta (ablation)DeltaEdit on Llama3-8B · CounterFact (3000 edits)efficacy_top / specificity_topHurt

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy