VibeWorlding：多模态智能体端到端构建 3D 开放世界的统一框架
原标题：Paper page - VibeWorlding: Can Multimodal Agents Construct 3D Open Worlds End-to-End?
AI 摘要
VibeWorlding 是一个用于基准测试和训练多模态智能体的统一框架，旨在从用户查询端到端构建交互式 3D 开放世界。该框架包含 VWE-BENCH 基准（含 2,616 个 3D 资产、323 个种子世界和 6,828 个查询）和 VibeWorlding-Gym 强化学习训练环境。实验表明，当前前沿多模态大模型（如 GPT-5.5 和 Qwen3.8-Max）的成功率低于 60%，而通过强化学习训练的开源模型 VibeWorlder-30B-A3B 在 Pass@1 上超越了闭源模型。
正文节选
VibeWorlding: Can Multimodal Agents Construct 3D Open Worlds End-to-End? Abstract A unified framework benchmarks and trains multimodal agents that infer intent, plan 3D scenes, invoke tools, and reflect on feedback, revealing that reinforcement learning improves open-source models beyond closed-source frontiers. Constructing an interactive 3D open world from a user query is important. However, existing methods are primarily evaluated on idealized, simple queries, making it difficult to systemati