AI generates natively editable PPTX from any document — real PowerPoint shapes, not images — no design skills needed · by Hugo He

[CVPR'24 Highlight] GPT4Point: A Unified Framework for Point-Language Understanding and Generation.

[ICLR 2025] "Understanding Constraint Inference in Safety-Critical Inverse Reinforcement Learning"

[CVPR 2025] VideoWorld is a simple generative model that learns purely from unlabeled videos—much like how babies learn by observing their environment.

PointWorld: Scaling 3D World Models for In-The-Wild Robotic Manipulation

Code to pretrain, fine-tune, and evaluate DreamZero and run sim & real-world evals

HY-Embodied: Embodied Foundation Models for Real-World Agents

ARIS ⚔️ (Auto-Research-In-Sleep) — Lightweight Markdown-only skills for autonomous ML research: cross-model review loops, idea discovery, and experiment automation. No framework, no lock-in — works…

你想蒸馏的下一个员工，何必是同事。蒸馏任何人的思维方式——心智模型、决策启发式、表达DNA。Distill how anyone thinks.

An agentic skills framework & software development methodology that works.

The agent harness performance optimization system. Skills, instincts, memory, security, and research-first development for Claude Code, Codex, Opencode, Cursor and beyond.

A set of ready to use Agent Skills for research, science, engineering, analysis, finance and writing.

Elevate your AI research writing, no more tedious polishing ✨

Universal skills loader for AI coding agents - npm i -g openskills

WorldArena: A Unified Benchmark for Evaluating Perception and Functional Utility of Embodied World Models

Official codebase for Fast-WAM: Do World Action Models Need Test-time Future Imagination?

[CVPR'26] Semi-Supervised Conformal Prediction With Unlabeled Nonconformity Score

Official code base for LeWorldModel: Stable End-to-End Joint-Embedding Predictive Architecture from Pixels

EVA: Aligning Video World Models with Executable Robot Actions via Inverse Dynamics Rewards

PAct: Part-Decomposed Single-View Articulated Object Generation

[RSS 2026] Causal video-action world model for generalist robot control

Advancing Open-source World Models

(arXiv) MixGRPO: Unlocking Flow-based GRPO Efficiency with Mixed ODE-SDE

Reference PyTorch implementation and models for DINOv3

code for "Diffusion Forcing: Next-token Prediction Meets Full-Sequence Diffusion"

[ICML 2025] Official PyTorch Implementation of "History-Guided Video Diffusion"

AnyPos: Automated Task-Agnostic Actions for Bimanual Manipulation

Official repo for vidar and vidarc: video foundation model for robotics.

WoW (World-Omniscient World Model) is a generative world model trained on 2 million robotic interaction trajectories, designed to imagine, reason, and act in the physical world. Unlike passive vide…