Amap Releases ABot-Earth 0.5: Shifting from 2D Distillation to 3D Native for Consistent Scene Generation
Original: 高德发布ABot-Earth0.5：跨越2D蒸馏模式，以3D原生驱动高一致性场景生成
Amap released ABot-Earth 0.5, a 3D native model bypassing 2D distillation to achieve highly consistent 3D scene generation.
Amap has released ABot-Earth 0.5, its latest spatial intelligence model. Moving beyond traditional 2D distillation methods (like Score Distillation Sampling), the model adopts a 3D native driving architecture. This breakthrough addresses multi-view inconsistency and distortion, enabling highly consistent 3D scene generation for autonomous driving simulation, smart cities, and digital twin mapping.
高德地圖（Amap）近日宣布推出其空間智能與 3D 場景生成模型 **ABot-Earth 0.5**。此版本核心突破在於捨棄了業界常見的「2D 圖像蒸餾至 3D」路徑，轉而採用「3D 原生驅動（3D Native）」架構，實現了高一致性、高逼真度的三維場景生成。
Free shows the 3-line summary; Pro unlocks the full deep summary (~300 words) so you never have to click through.
See Pro plans →Want the original English / full article?
Read on 量子位 QbitAI →Summaries are AI-generated; the original article is authoritative.