AI Paper Atlas

Research map / Large language models / Retrieval and knowledge

Knowledge Editing with Subspace-Aware Key-Value Mappings

Haewon Park, Sangwoo Kim, Yohan Jo

ICLR 2026 · Large language models › Retrieval and knowledge · approach: Knowledge editing in LLMs

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
SUIT instead of AlphaEditLLaMA3-8B-Instruct · CounterFact batch 100Specificity: 31.0 -> 74.2Improved
SUIT instead of AlphaEditGPT-J-6B · CounterFact batch 100S: 73.0 -> 84.7Improved
SUIT instead of AlphaEdit at batch size 1LLaMA3-8B-Instruct · CounterFact batch 1S: 36.6 -> 85.4Improved
SUIT instead of AlphaEdit on zsREQwen2.5-7B-Instruct · zsRE Spe. l=1Specificity: 30.1 -> 64.4Improved
SUIT instead of AlphaEditLLaMA3-8B-Instruct · General Capability after 1000 editsGC: 62.2 -> 63.0No clear effect
add subspace-aware key vector k' only (ablation)AlphaEdit baseline · CounterFact LLaMA3S: 55.8 -> 82.2Improved
add subspace-aware residual vector delta' only (ablation)AlphaEdit baseline · CounterFact LLaMA3S: 55.8 -> 68.3Improved
combine k' and delta' (full SUIT) vs k' only (ablation)SUIT · CounterFact LLaMA3S: 82.2 -> 86.8Improved
add delta' to k' (ablation)SUIT · CounterFact LLaMA3Generalization: 77.9 -> 90.3Improved
raise energy threshold tau_energy (ablation)SUIT · CounterFact LLaMA3Specificity vs EfficacyMixed
raise directional penalty lambda (ablation)SUIT · CounterFact LLaMA3Specificity vs GeneralizationMixed
project edit onto 2D critical subspace span(w1,w2) vs orthogonal remainder (ablation)SUIT residual vector · CounterFact 1000 editsp(a*): 0.59 -> 0.67Improved

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