AI Paper Atlas

Research map / Large language models / Retrieval and knowledge

RGMem: Renormalization Group–inspired Memory Evolution for Language Agents

Ao Tian, Yunfeng Lu, Xinxin Fan, Changhao Wang, Lanzhi Zhou, Yeyao Zhang, Yanfang Liu

ICML 2026 · Large language models › Retrieval and knowledge · approach: Retrieval-augmented reasoning · 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
Use RGMem multi-scale memory instead of best baseline (Memory OS)GPT-4.1 backbone · PersonaMem (128k)average accuracy (%): 65.03 -> 74.01Improved
Use RGMem instead of best baseline (Memory OS)GPT-4o-mini backbone · PersonaMem (128k)average accuracy (%): 56.79 -> 63.87Improved
Use RGMem instead of Zepgpt-4o-mini · LOCOMOaverage accuracy (%): 75.14 -> 78.92Improved
Use RGMem instead of Full-Contextgpt-4.1-mini · LOCOMOaverage accuracy (%): 87.52 vs 86.17No clear effect
Use RGMem instead of Full-Contextgpt-4o-mini · LOCOMOaverage accuracy (%): 71.41 -> 78.92Improved
Increase retrieved context from about 3k to beyond 3.8k tokens (ablation)RGMem · LOCOMOaccuracyMixed
Move evolution threshold away from 3 (both lower and higher) (ablation)RGMem · LOCOMOaccuracyHurt
Move evolution threshold away from 3 (ablation)RGMem · PersonaMem Latest PreferenceaccuracyHurt
Use RGMem instead of Memory OS on open-domain questionsgpt-4o-mini · LOCOMO Openaccuracy (%): 67.71 vs 67.71No clear effect

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