Physical AI News (Japan)

Physical AI, robotics and compute — reported from Japan.

日本語
· Research & technology· 日本語で読む

SimpleMemVLA study linked to ModelBest uses visual history as memory

SimpleMemVLA was presented as a method for robots that uses timestamped visual history directly as context. The article reported good results on several memory benchmarks. The study was described as a joint proposal by ModelBest and others.

What happened

Leaderobot.com published an article introducing SimpleMemVLA on September 28, 2026. The article described it as a method that uses timestamped visual history as context without adding a dedicated memory module.

Sources

Translated from the Japanese original: /news/menpeki-chinou-kanren-no-simplememvla-shikaku-rireki-wosonom-c76cfe. The Japanese version is authoritative if they differ.