AI Paper Atlas

Research map / Large language models / Retrieval and knowledge

SR-KI: Scalable and Real-Time Knowledge Integration into LLMs via Supervised Attention

Bohan Yu, Wei Huang, Kang Liu

AAAI 2026 · Large language models › Retrieval and knowledge · approach: Knowledge graph reasoning with LLMs · 2 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 supervised attention training and top-k KB selection (SR-KI) instead of KBLaMKBLaM-style KV injection · KB size 10000 QAreference ID accuracy: 0.0087 -> 0.7800Improved
use SR-KI instead of KBLaM at small KB sizeKB size 100 · knowledge BERTScore F1BERTScore F1: 0.8725 -> 0.8547Hurt
reuse retrieval-layer top-k indices in subsequent layers (ablation)SR-KI · KB size 40000reference ID accuracy: 0.3600 -> 0.6940Improved
reuse retrieval-layer top-k indices in subsequent layers (ablation)SR-KI · KB size 1000reference ID accuracy: 0.8167 -> 0.9467Improved
add supervised attention loss at the retrieval layerSR-KI · KB retrieval, KB size 1000Recall@100: 0.4375 -> 0.9975Improved
add supervised attention trainingSR-KI · unanswerable QA refusal accuracyrefusal accuracyHurt
use SR-KI instead of ICL at 100 KBsin-context learning · KB size 100reference ID accuracy: 0.6730 -> 0.9837Improved
use SR-KI instead of KBLaM on alias-perturbed queriesKBLaM · generalization, KB size 1000ID generalization accuracy: 0.6810 -> 0.8089Improved
scale KB size from 1000 to 40000SR-KI · QA reference ID accuracyreference ID accuracy: 0.9467 -> 0.6940Hurt
choose layer 25 as retrieval layer by single-layer injection (ablation)Qwen2.5-7B-Instruct · layer identification (100 samples)ID accuracy and EM accuracyImproved

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