A free-flyer rebuilt the ISS interior from photos it took five years ago

Bernardo Taveira and colleagues at Chalmers University of Technology and Zenseact have built VoroTracing, a differentiable renderer that traces camera rays cell by cell through a Voronoi tessellation and reaches 623 frames per second on Mip-NeRF 360 — 2.8× the throughput of 3D Gaussian Splatting and 3.2× that of the fastest previous ray-based method.

Why it matters: Rasterisation is fast because it assumes a pinhole camera. Anything that breaks that assumption — barrel distortion, a rolling shutter, motion blur, a shallow depth of field — has to be bolted on as a special case, and each one is its own engineering project.

Zhiqiang Hu, Shouren Huang and Masatoshi Ishikawa at Tokyo University of Science have built GaussMemory, which swaps the 2D token buffers that vision-language-action models normally use for a persistent 3D gaussian scene memory — and trains the policy to work out which gaussians are worth tracking carefully.

The problem: Put an apple in one of two identical drawers. Close it. Go and do something else. Come back and retrieve the apple.

Hao Qin and colleagues at Zhejiang University have published NGS-Marker at ICLR 2026, a watermarking scheme that embeds an ownership message into gaussian primitives themselves rather than into the images they render.

Why it matters: A splat is an explicit list of primitives, which is most of what makes it convenient — and it means anyone can open a scene, select a region, and paste it into their own. The paper calls this partial infringement, and the example it opens with is a polar bear lifted out of one capture and dropped onto someone else's park bench.

Quan-Dung Pham and colleagues at VinMotion and the University of Southern California have built HumanoidVLN, a physics-grounded simulator whose environments are drawn partly from gaussian splatting reconstructions of real interiors — and whose headline result is how badly current navigation models cope with having legs.

Why it matters: Vision-language navigation benchmarks have mostly assumed a wheeled robot gliding along a floor plan. A bipedal robot has to stay upright, its camera pitches and rolls with every step, and no two humanoid platforms are shaped alike.

Wei Zhang and colleagues at Northwestern Polytechnical University and vivo's BlueImage Lab identify a specific failure in how city-scale gaussian splats are built: each level of the octree is optimised in isolation, with no communication between them.

Why it matters: Octree-based anchor splatting is how the field currently scales to whole cities — coarse levels hold building masses, fine levels hold detail. It works, and the artefacts it produces have mostly been treated as the price of scale.