Tag
WorldRuler 是一个基于可执行 3D 规则图的生成式世界模型,能在保存与加载时持久化用户自定义规则与世界状态,并通过闭环的规则执行、视频生成与感知反馈准确实现规则引发的视觉变化。作者还基于 GTA-V 和 Habitat 构建了 RuleVerse 数据集以支持规则驱动交互的研究。
Google DeepMind 发布了 Genie 4 世界模型研究预览版,新增实时音频生成、更逼真的可交互世界,以及支持中途提示词编辑的实时操控功能。该模型可维持最长两分钟的场景一致性,并与 Gemini 深度集成,用于机器人仿真、游戏和影视原型制作等场景。
网友发现了一个名为 WeduLab 的宝藏网站,将物理、化学、生物、数学、地理等学科实验直接搬进浏览器,无需安装即可动手进行电路搭建、3D等高线、大气环流、水循环、细胞结构、DNA组装、分子构型等交互式模拟实验。
RAR (Rate-Adaptive Rollout) 是一个让交互式世界模型以可控时间速率进行 rollout 的框架,通过 Stride-RoPE 将不同采样速率的潜在帧对齐到共享物理时间坐标,实现多速率生成、未见速率泛化与在线速率切换,同时用多步长视频采样兼顾细粒度动态与长时程演化。
@jinglian 介绍并推荐“交互式教学工具集”WeduLab 的最新更新:新增 3 个工具(含 K12 化学元素周期表),累计已达 25 个开箱即用的 3D 交互式学科实验工具,覆盖物理、地理、生物、化学、数学等学科,作者 Teacher WY 计划在 Claude 可用的情况下持续更新。
Extend engineer Bo Lau built a highly detailed interactive website that simulates sitting at Elizabeth Holmes' desk, featuring real emails, texts, slides, and documents sourced from the United States v. Holmes criminal trial exhibits, including a working replica of the Theranos Edison machine.
WIRED published an interactive web feature, 'Spend Elon Musk's Money,' letting users virtually allocate the tech billionaire's fortune across purchases and donations — a lighthearted viral simulation.
论文提出 Trace2Env,一个免学习的 agentic 语言世界模型框架,通过从历史交互轨迹重建环境世界书(world book)来模拟原始系统不可用的交互环境。在九个环境上的实验表明,其在下一观测保真度和长程交互一致性方面优于基于提示的传统语言世界模型。
HelixWorld is a real-time interactive audio-visual world model where visual scenes and camera-grounded spatial stereo sound co-evolve under user interaction, using a teacher-student distillation pipeline to run at 24 FPS on a single GPU. The paper also introduces HelixBench to evaluate spatial-acoustic consistency of synthesized sound fields.
The tweet describes using Claude Opus 5.5 to build an interactive Raptor 3 rocket engine simulation in the browser, allowing users to explore how rocket engines work through hands-on interaction.
Magpie is a real-time generative rendering system for interactive games that separates gameplay logic from visual generation to reduce asset production costs and preserve interactive designability.
DreamForge-World 0.1 Preview is a low-compute world model that enables real-time interactive simulation on consumer GPUs, supporting keyboard/mouse control and achieving 14-15 FPS at 480p resolution on a single RTX 4090.
An interactive Alpha Stirling engine simulation built with ChatGPT and Three.js featuring animated kinematics, heat-driven motion, and particle gas dynamics.
γ-World is a generative multi-agent world model that supports independently controllable, permutation-symmetric agents using Simplex Rotary Agent Encoding and Sparse Hub Attention, achieving real-time 24 FPS rollouts and zero-shot generalization from two to four players.
OdysseyML introduces Starchild-1, the first real-time multimodal world model that can generate interactive simulations of the world with audio, marking a significant step toward a general-purpose world simulator.
Overworld releases Waypoint-1.5, a real-time video world model designed for everyday GPUs, featuring improved visual fidelity and new 360p and 720p tiers for broader hardware accessibility.
Google has launched Project Genie, an experimental prototype for Google AI Ultra subscribers that allows users to create, explore, and remix infinite interactive worlds using Genie 3.
Google DeepMind 发布 Genie 4,这是其迄今最真实、最可交互的世界模型,基于 Gemini 构建,支持实时交互、键盘导航与自然语言控制场景,正向可信测试者和 AI Ultra 用户推出研究预览。