AI Paper Atlas

Research map / Vision-language and multimodal / Human and embodied modeling

Task-Aware Structured Memory for Dynamic Multi-modal In-Context Learning

Zhirui Chen, Ziwei Chen, Ling Shao

ICML 2026 · Vision-language and multimodal › Human and embodied modeling · approach: Multimodal retrieval-augmented modeling · 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
task-vector scoring instead of hard pruning (ablation)hard-pruning baseline, 20% budget · ImageNet-100accuracy: 45.3 -> 55.2Improved
add semantics-aware token merging (ablation)hard pruning + task vector scoring · ScreenSpotscore: 12.1 -> 18.8Improved
add dynamic retrieval from latent bank (ablation)pruning + task vector + merging · YouCook2score: 98.5 -> 109.5Improved
layer-adaptive gating instead of uniform lambda 0.5 (ablation)TASM · ScreenSpotscore: 16.2 -> 19.5Improved
task-vector scoring instead of answer-aware attention (ablation)EMLoC · average over IN-100, MME, OK-VQAaverage score: 54.8 -> 56.2Improved
task-vector scoring instead of attention (H2O) (ablation)H2O scoring · average over IN-100, MME, OK-VQAaverage score: 45.1 -> 56.2Improved
JS-divergence trigger instead of periodic-dense retrieval (ablation)periodic-dense retrieval · YouCook2score: 108.9 -> 109.5Improved
token merging instead of hard pruning (EMLoC) at 10% retention (ablation)EMLoC · ScreenSpotscore: 12.5 -> 15.9Improved
transfer task vector from source classes to disjoint target classes (ablation)TASM oracle · ImageNet-100accuracy: 65.0 -> 64.2No clear effect
replace task vector with random projection (ablation)TASM · ImageNet-100accuracy: 65.0 -> 15.3Hurt
TASM instead of EMLoCEMLoC · V-NIAH 200 imagesaccuracy: 18.9 -> 45.5Improved
TASM instead of EMLoCEMLoC · Qwen2-VL-7B efficiency at 32kGPU memory (GB): 10.2 -> 11.5Hurt

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