This survey revisits spatial intelligence research through the lens of cognitive maps, arguing that agents must build a stable, updatable, and inferable internal representation to integrate partial observations into a coherent world model for perception, reasoning, and generation.
PAT3D, a Physics‑Augmented Text‑to‑3D scene generation framework presented at ICLR 2026, extracts object‑space relationships from text‑driven images, initializes a hierarchical layout, and refines it with differentiable rigid‑body simulation and semantic loss, yielding physically stable, editable scenes that outperform prior methods in stability metrics and enable downstream editing, animation, and robot simulation.
PAT3D, a physics‑augmented text‑to‑3D scene generation framework presented at ICLR 2026, extracts object relationships, initializes a hierarchical layout, and optimizes it via differentiable rigid‑body simulation, producing physically stable, semantically faithful scenes that can be directly edited, animated, or used for robot simulation, outperforming prior methods.
This article explores how Alibaba’s 3D scene‑based recommendation system combines computer‑vision, deep‑learning and data‑driven matching algorithms to create immersive, size‑accurate product visualizations that enhance user experience and drive higher click‑through rates in e‑commerce.
This article explores how AI is used in iQIYI's video advertising pipeline to analyze video content, generate and recommend ad placement points, create scene‑aware ad creatives, build a video knowledge graph, and support various ad formats, ultimately improving ad relevance and revenue.
