TIPSv2 (CVPR'26) and TIPS (ICLR'25)

Cambrian-S: Towards Spatial Supersensing in Video

[CVPR 2026] G2VLM: Geometry Grounded Vision Language Model with Unified 3D Reconstruction and Spatial Reasoning

SpatialClaw: Rethinking Action Interface for Agentic Spatial Reasoning

Visual Spatial Tuning

[CVPR 2026] Official release of "Spatial-SSRL: Enhancing Spatial Understanding via Self-Supervised Reinforcement Learning"

[ICCV 2025] Enhancing spatial understanding in text-to-Image diffusion models

[NeurIPS DB 2025] IR3D-Bench: Evaluating Vision-Language Model Scene Understanding as Agentic Inverse Rendering

[CVPR 2026 (Highlight)] 4D-RGPT: Toward Region-level 4D Understanding via Perceptual Distillation

[ICLR 2026] 🦅 FALCON: an effective vision-language-action model injects rich 3D spatial tokens into the action head, enabling robust spatial understanding and SOTA performance across diverse manipulation tasks.

[AAAI 2026 Oral] STRIDE-QA: Visual Question Answering Dataset for Spatiotemporal Reasoning in Urban Driving Scenes

POMA-3D: The Point Map Way to 3D Scene Understanding.

[ECCV 2026] One Scene, Two Depths: Probing Geometric Ambiguity in Monocular Foundation Models (Layered 3D Spatial Understanding)

SIFThinker: Spatially-Aware Image Focus for Visual Reasoning

[CVPR 2026] HandVQA: Diagnosing and Improving Fine-Grained Spatial Reasoning about Hands in Vision-Language Models

Autoregressive Mosaics is a project that attempts to force an LLM trained only on text to paint a picture one discrete pixel at a time.