AI Paper Atlas

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

MemoryVLA: Perceptual-Cognitive Memory in Vision-Language-Action Models for Robotic Manipulation

Hao Shi, Bin Xie, Yingfei Liu, Lin Sun, Fengrong Liu, Tiancai Wang, Erjin Zhou, Haoqiang Fan, Xiangyu Zhang, Gao Huang

ICLR 2026 · Vision-language and multimodal › Robot action and navigation · approach: Vision-language-action foundation models

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 both perceptual and cognitive memory instead of cognitive only (ablation)MemoryVLA · SimplerEnv-Bridgeavg success rate (%): 63.5 -> 71.9Improved
use both memories instead of perceptual only (ablation)MemoryVLA · SimplerEnv-Bridgeavg success rate (%): 64.6 -> 71.9Improved
set memory length to 16 instead of 4 (ablation)MemoryVLA · SimplerEnv-Bridgeavg success rate (%): 67.7 -> 71.9Improved
set memory length to 16 instead of 64 (ablation)MemoryVLA · SimplerEnv-Bridgeavg success rate (%): 67.7 -> 71.9Improved
add timestep positional encoding to retrieval (ablation)MemoryVLA · SimplerEnv-Bridgeavg success rate (%): 69.8 -> 71.9Improved
use gate fusion instead of simple addition (ablation)MemoryVLA · SimplerEnv-Bridgeavg success rate (%): 67.7 -> 71.9Improved
use token-merge consolidation instead of FIFO (ablation)MemoryVLA · SimplerEnv-Bridgeavg success rate (%): 66.7 -> 71.9Improved
add perceptual-cognitive memory (MemoryVLA over CogACT)MemoryVLA · Real-world long-horizon temporal tasksavg success score (%): 57 -> 83Improved
add perceptual-cognitive memory (MemoryVLA over CogACT)MemoryVLA · Mikasa-Roboavg success rate (%): 29.4 -> 41.2Improved
unseen camera view (distribution shift)MemoryVLA · Pick Coke Can (sim OOD)success rate (%): 92.0 -> 42.0Hurt

Later papers that compared with this method

8 later accepted papers report MemoryVLA as a baseline in their results tables; 8 of them report a better number on most of the results they share with it.

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