Surflo: global-state surface generation from unposed views

What changed

Compressed variable numbers of unposed RGB views into fixed-size global latent tokens and decoded oriented surface points using conditional flow matching, with photometric/depth guidance to reduce local inconsistency.

Scientific scope

Compressed variable numbers of unposed RGB views into fixed-size global latent tokens and decoded oriented surface points using conditional flow matching, with photometric/depth guidance to reduce local inconsistency. Inherits frozen VGGT features and flow matching. Arbitrarily chosen output point counts do not imply arbitrary recovered geometric detail, exact hidden geometry or constant decoding cost; guidance adds inference computation.

Evidence

The linked original sources describe the stated contribution and its evidence limits. WRS has not independently rerun the experiments or certified a proof. Inherits frozen VGGT features and flow matching. Arbitrarily chosen output point counts do not imply arbitrary recovered geometric detail, exact hidden geometry or constant decoding cost; guidance adds inference computation.

Contributors

• Source-listed authors and technical contributors · SCIENCE: Source-listed authors or explicitly credited technical contributors for the cited versions. Family entries retain source-specific attribution; a combined list is not one original-paper byline or equal individual work. Names alone do not establish resolved identities.

• Surflo: Consistent 3D Surface Flow Model with Global State
• Kyutai scientific publications index

Provisional WRS assessment, edition 2026-10-16.1. Independent expert calibration is pending. This grade does not certify correctness or assign ownership.

Full scientific record · Browse contribution grades · Request a correction