From 2a953ba997c19a150f10e0f38ea1902d88c83a60 Mon Sep 17 00:00:00 2001 From: Ceither Date: Wed, 8 Apr 2026 17:31:29 +0800 Subject: [PATCH] Add LeWorldModel documentation Introduced LeWorldModel, a stable end-to-end JEPA from raw pixels with reduced loss hyperparameters. --- app/docs/CommunityShare/Geek/leworldmodel.md | 13 +++++++++++++ 1 file changed, 13 insertions(+) create mode 100644 app/docs/CommunityShare/Geek/leworldmodel.md diff --git a/app/docs/CommunityShare/Geek/leworldmodel.md b/app/docs/CommunityShare/Geek/leworldmodel.md new file mode 100644 index 00000000..efdaac94 --- /dev/null +++ b/app/docs/CommunityShare/Geek/leworldmodel.md @@ -0,0 +1,13 @@ +--- +title: "LeWorldModel" +description: "Stable End-to-End Joint-Embedding Predictive Architecture from Pixels" +date: "2026-04-08" +tags: + - "世界模型" +--- + +Joint Embedding Predictive Architectures (JEPAs) offer a compelling framework for learning world models in compact latent spaces, yet existing methods remain fragile, relying on complex multi-term losses, exponential moving averages, pretrained encoders, or auxiliary supervision to avoid representation collapse. In this work, we introduce LeWorldModel (LeWM), the first JEPA that trains stably end-to-end from raw pixels using only two loss terms: a next-embedding prediction loss and a regularizer enforcing Gaussian-distributed latent embeddings. This reduces tunable loss hyperparameters from six to one compared to the only existing end-to-end alternative. With 15M parameters trainable on a single GPU in a few hours, LeWM plans up to 48× faster than foundation-model-based world models while remaining competitive across diverse 2D and 3D control tasks. Beyond control, we show that LeWM’s latent space encodes meaningful physical structure through probing of physical quantities. Surprise evaluation confirms that the model reliably detects physically implausible events. + +
+ +