Skip to content
Agientry
← Back to Videos

RAVEN Stores Long-Horizon Robot Memories as Visual Embeddings

Loading video
Loading video

RAVEN stores visual embeddings instead of captions, allowing a robot to recall fine object details across days and navigate to goals without losing information at a text bottleneck.

Robot MemoryVisual EmbeddingsNavigationLong-Horizon AutonomyRAVEN
Category: research
Author: @yixun_hu
Date: 2026-11-02T00:00:00
Duration: 23.659s
Source: https://x.com/yixun_hu/status/2117310956711620737
Reference: https://ravenmem.github.io

Related Videos

SceneBloom: Simulation-Ready Scenes for Embodied Learning
0:38

SceneBloom: Simulation-Ready Scenes for Embodied Learning

@l_yuhan7272 · 110 views · 2026-10-28
SceneBloomEmbodied AI
AGC-VLN: Air-Ground VLN via Shared Bird's-Eye Maps
1:59

AGC-VLN: Air-Ground VLN via Shared Bird's-Eye Maps

@heetezition · 595 views · 2026-09-07
VLNAir-Ground Collaboration
3D World Simulator Runs Interactive Robot Demos in Browser
0:47

3D World Simulator Runs Interactive Robot Demos in Browser

@YunzhuLiYZ · 23 views · 2026-11-03
World Models3D Interaction
DynaSolver Simulates Fluid Around Moving Robots
1:16

DynaSolver Simulates Fluid Around Moving Robots

@HaoZhao_AIRSUN · 18 views · 2026-11-03
Fluid SimulationNeural Operator
3D World Simulator Predicts Geometry and Touch
0:11

3D World Simulator Predicts Geometry and Touch

@robotsdigest · 20 views · 2026-11-03
World Models3D Scene Prediction
Ming 3.0 Opens a Full-Stack Audio Stack
0:56

Ming 3.0 Opens a Full-Stack Audio Stack

@Lonely__MH · 16 views · 2026-11-03
Speech & AudioTTS

As an Amazon Associate, we earn from qualifying purchases.