AI Paper Atlas

Research map / Vision-language and multimodal / Robot action and navigation

CURE: Cumulative Knowledge Reuse for Efficient Device-Server Hybrid Inference in Vision-Language Models

Youngjun Lee, Doyoung Kim, Junhyeok Kang, Hwanjun Song, Jae-Gil Lee

ECCV 2026 · Vision-language and multimodal › Robot action and navigation · approach: Vision-language-action foundation models · 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 server-side predictions to the reservoir update (ablation)CURE · 13 datasets (avg)top-1 accuracy (%): 78.26 -> 85.75Improved
add server-side predictions to the reservoir update (ablation)CURE · 13 datasets (avg)delegation frequency (%): 37.15 -> 33.26Improved
add on-device confident predictions to the reservoir update (ablation)CURE · 13 datasets (avg)delegation frequency (%): 39.55 -> 33.26Improved
add on-device confident predictions to the reservoir update (ablation)CURE · 13 datasets (avg)top-1 accuracy (%): 85.48 -> 85.75Improved
add knowledge fusion of VLM and reservoir predictions (ablation)CURE · 13 datasets (avg)top-1 accuracy (%): 83.01 -> 85.75Improved
use shared covariance instead of class-wise covariance in the reservoir (ablation)CURE · 13 datasets (avg)top-1 accuracy (%): 85.75 -> 85.71No clear effect
use shared covariance instead of class-wise covariance in the reservoir (ablation)CURE · 13 datasets (avg)delegation frequency (%): 33.26 -> 33.51No clear effect
add cumulative knowledge reuse (reservoir) over non-cumulative hybridNon-Cumulative Hybrid · 13 datasets (avg)delegation frequency (%): 74.92 -> 33.26Improved
add cumulative knowledge reuse over always-delegateAlways-Delegate · 13 datasets (avg)top-1 accuracy (%): 84.70 -> 85.75No clear effect
add cumulative knowledge reuse over non-cumulative hybridNon-Cumulative Hybrid · Stanford Carsend-to-end runtime (s): 1848.71 -> 659.64Improved
add cumulative knowledge reuse over always-delegateAlways-Delegate · Stanford Carsend-to-end runtime (s): 2421.73 -> 659.64Improved
use device-server delegation instead of on-device-only test-time adaptationDOTA · 13 datasets (avg)top-1 accuracy (%): 80.63 -> 85.75Improved

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