AI Paper Atlas

Research map / Large language models / Retrieval and knowledge

MedREK: Retrieval-Based Editing for Medical LLMs with Key-Aware Prompts

Shujun Xia, Haokun Lin, Yichen Wu, Yinan Zhou, Zixuan Li, Zhongwei Wan, Xingrun Xing, Yefeng Zheng, Xiang Li, Caifeng Shan, Zhenan Sun, Quanzheng Li

ICML 2026 · Large language models › Retrieval and knowledge · approach: Knowledge-augmented question answering · 0 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 shared query-key MLP (ablation)MedREK · MedCF++ 100 edits Meditron-7BAvg: 69.91 -> 88.30Improved
add shared query-key MLP (ablation)MedREK · MedCF++ 100 edits Meditron-7BEfficacy: 56.74 -> 77.96Improved
add attention-based prompt encoder (ablation)MedREK · MedCF++ 100 edits Meditron-7BAvg: 85.24 -> 88.30Improved
add attention-based prompt encoder (ablation)MedREK · MedCF++ 100 edits Meditron-7BGenerality: 75.63 -> 79.88Improved
use MedREK instead of RECIPE under 100-edit batchretrieval-based editor · MedCF++ Meditron-7BAvg: 71.83 -> 88.30Improved
use MedREK instead of RECIPE under 100-edit batchretrieval-based editor · MedCF++ Meditron-7BEfficacy: 68.30 -> 77.96Improved
use MedREK instead of RECIPE under 100-edit batchretrieval-based editor · MedCF++ Meditron-7BGenerality: 70.74 -> 79.88Improved
increase batch size from 1 to 100 editsRECIPE · MedCF++ Meditron-7BAvg: 84.11 -> 71.83Hurt
increase batch size from 1 to 100 editsMedREK · MedCF++ Meditron-7BAvg: 89.33 -> 88.30No clear effect
use retrieval-based MedREK instead of parameter-modifying MedLaSAeditor · MedVersa single edit Meditron-7BAvg: 79.97 -> 86.24Improved
use MedREK instead of MEMIT under 100 editseditor · MedCF++ HuatuoGPT-o1-8BAvg: 61.23 -> 86.83Improved
use MedREK instead of RECIPE on MedVersa single editeditor · MedVersa Meditron-7BAvg: 78.35 -> 86.24Improved

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