Skip to content
Agientry
← Back to Videos

AgentGarten: Code Worlds for Evolving Agents

Loading video
Loading video

AgentGarten couples simulators and game engines with a shared neural renderer: a code engine keeps persistent state and runs program-defined rules while a real-time diffusion renderer turns structured conditions into what the agent sees. Agents learn tasks in about 4 rounds versus millions for conventional RL, distilling each round into playbooks that later agents inherit and refine.

World ModelsAgentSimulated EnvironmentNeural RendererReinforcement LearningMirroS-Lab
Category: world-model
Author: @JiangranLyu
Date: 2026-10-09T00:00:00
Duration: 8.0s
Source: https://x.com/JiangranLyu/status/2108452941766049868
Reference: https://arxiv.org/abs/2610.12374

Related Videos

LLM 3D Worlds Arena: 14 models, one prompt, blind votes

LLM 3D Worlds Arena: 14 models, one prompt, blind votes

@halcyonrayes · 23 views · 2026-10-12
3D GenerationWorld ModelsLeaderboard
Odyssey 3: World Models Let Robots Rehearse Physics
0:29

Odyssey 3: World Models Let Robots Rehearse Physics

@joaquintenest · 173 views · 2026-10-09
World ModelsSim-to-RealRobot training
XPENG XPACE: A Unified Embodied World Model for IRON
1:31

XPENG XPACE: A Unified Embodied World Model for IRON

@CyberRobooo · 629 views · 2026-09-17
World ModelsEmbodied AIXpeng
PointZero: 3D World Model Pretrained Without Robot Data
0:43

PointZero: 3D World Model Pretrained Without Robot Data

@bardienus · 534 views · 2026-09-17
World Models3D DynamicsPoint Tracking
OpenWAM open world-action model pretraining framework
3:00

OpenWAM open world-action model pretraining framework

@KnightNemo_ · 562 views · 2026-09-09
World ModelsRobot LearningPretraining
Programmable World Model: Executable State Separated From Rendering
1:17

Programmable World Model: Executable State Separated From Rendering

@alayastd · 695 views · 2026-09-09
World ModelsVideo GenerationEmbodied AI

As an Amazon Associate, we earn from qualifying purchases.