HuggingPapers· @HuggingPapers · X·· 6 小时前AI 评分
Act3D：VLM用RL探索3D场景
Act3D 学会在 3D 中思考和探索 给 VLM 一张图像，然后用仅结果的 RL 决定何时以及如何移动相机并收集证据。在六个 3D 基准上平均 53.2%，比其 Qwen 基座高 +22%，超过 Gemini 3.6 Flash。
Act3D learns to think and explore in 3D
A VLM is given one image, then uses outcome-only RL to decide when and how to move the camera and gather evidence. 53.2% average across six 3D benchmarks, +22% over its Qwen base and above Gemini 3.6 Flash.
来源：HuggingPapers · x.com