World models are neural networks that learn an internal representation of an environment’s dynamics, enabling agents to simulate and plan within a learned latent space. They are used in model-based reinforcement learning, robotics, and autonomous decision-making.
Logic Language for World Models 🌱🐋🌍
Joint Embedding Predictive Architecture for World Models, written in Rust.
MosaicMem in Rust - Hybrid Spatial Memory for Video World Models
Pure-Rust reproduction and extension of LeWorldModel (Maes et al., 2026).
A commit-and-audit proof system for deterministic, quantized inference of a JEPA-style world model (LeWorldModel)
Implementation of Generative Robot Policies via Predictive World Modeling in Rust
The latest generation of Claude is Anthropic's state-of-the-art AI model series (headlined by Claude 5 / Claude Opus 5 and Claude Sonnet 5), engineered for advanced reasoning, agentic coding, multi-step problem solving, and complex enterprise workflows with an expanded 1-million-token context window.
Inspect and compare self-supervised vision model representations for DINOv2, I-JEPA, V-JEPA 2, EUPE
Persistent geospatial world models.
High-performance LeRobot dataloader and replay buffer with Rust-powered video decoding, zero-copy tensors, and optional NVDEC GPU acceleration.
Minimal native-Rust proof of durable generative-state reuse: tiny ConvLSTM state persisted through EntropyFS across process death (VOLE-Field, DOI 10.5281/zenodo.22805773).
A grammar and method for progressively resolved worlds, concepts, understanding, and stories, with a Rust engine and MCP server.
Energy First Architecture (EFA): a post-transformer, on-device, energy-based architecture for Physical AI. Charlot Lab · Institute for Physical AI @ BMI.