Mar 29, 2026 · Artificial Intelligence

How LoGeR Enables Minute‑Long 3D Reconstruction with Hybrid Memory

The article presents LoGeR, a long‑context geometric reconstruction framework that combines test‑time‑training memory and sliding‑window attention to achieve minute‑scale, fully‑feedforward 3D reconstruction with superior accuracy on benchmarks such as KITTI and VBR.

3D ReconstructionComputer VisionHybrid Memory

0 likes · 11 min read

How LoGeR Enables Minute‑Long 3D Reconstruction with Hybrid Memory

AI Frontier Lectures

Mar 16, 2026 · Artificial Intelligence

How LoGeR Extends 3D Reconstruction to Thousands of Frames with Hybrid Memory

LoGeR, a new long‑context geometric reconstruction framework from DeepMind and UC Berkeley, uses a hybrid memory module combining test‑time‑training (TTT) and sliding‑window attention (SWA) to enable feed‑forward 3D reconstruction over sequences of up to tens of thousands of frames, achieving state‑of‑the‑art accuracy on KITTI, VBR, 7‑Scenes, ScanNetV2 and TUM‑Dynamics benchmarks.

3D ReconstructionHybrid MemoryLoGeR

0 likes · 11 min read

How LoGeR Extends 3D Reconstruction to Thousands of Frames with Hybrid Memory