AI Paper Atlas

Research map / Vision-language and multimodal / Multimodal LLMs

CoMem: Compositional Concept-Graph Memory for Vision–Language Adaptation

Heng Zhou, Jing Tang, Jusheng zhang, Yanshu Li, Canran Xiao, Liwei Hou, Zong Ke, Jiawei Yao

ICLR 2026 · Vision-language and multimodal › Multimodal LLMs · approach: Chain-of-thought multimodal reasoning

Read the paper

AI-extracted summary. Written by AI from the paper's full text. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Proposes CoMem, a compositional concept-graph memory for continual vision-language adaptation that avoids storing raw data under privacy and memory budgets.

Problem

Continual vision-language learning from non-stationary streams under privacy and memory constraints, where naive finetuning causes forgetting and harms transfer.

Approach

Uses a compositional concept-graph memory in place of raw-data replay to sustain stable yet plastic vision-language capability; details beyond the truncated abstract are unavailable.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
Use relation-aware MMD for the feature generator instead of vanilla RBF (ablation)CoMem · Cross-domain retrievalAvg mR: 75.7 -> 76.6Improved
Add entropy gate to replay distillation (ablation)CoMem · Cross-domain retrievalAverage forgetting (AF): 2.8 -> 1.9Improved
Add compositional consistency loss (PoE + relation contrast) (ablation)CoMem · Cross-domain retrievalAvg mR: 74.9 -> 76.6Improved
Use stop-gradient so replay loss does not update the generator (ablation)CoMem · Cross-domain retrievalAverage forgetting (AF): 2.6 -> 1.9Improved
Sample subgraphs with k-DPP + Steiner + MH instead of uniform (ablation)CoMem · Cross-domain retrievalAvg mR: 75.2 -> 76.6Improved
Add Metropolis-Hastings accept step after k-DPP + Steiner (ablation)CoMem · Cross-domain retrievalAvg mR: 76.2 -> 76.6Improved
Use frozen teacher with shared low-rank verifier instead of student verifier or dense per-concept projections (ablation)CoMem · Cross-domain retrievalAverage forgetting (AF): 2.5 -> 1.9Improved
Add edge anchors to node anchors in replay (ablation)CoMem · SVLC relation families and VQA relation skillsAUROC / accuracy: +1.7 pp AUROC, +1.6 pp AccImproved
Use fixed relation schema instead of dynamic relation schema (ablation)CoMem · Cross-domain retrievalAvg mR: 73.4 -> 76.6Improved
Use manually curated data instead of automated parsing (ablation)CoMem · Cross-domain retrievalAvg mR: 76.6 -> 76.2No clear effect
Increase anchor budget from 8K to 64K tokens (ablation)CoMem · Cross-domain retrievalAvg mR: 75.8 -> 76.7Improved
Scale backbone from ViT-B/16 to ViT-L/14 (ablation)CoMem · Cross-domain retrievalAvg mR: +0.6 to +1.1 ppImproved

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