XSPLAIN: XAI-enabling Splat-based Prototype Learning for Attribute-aware INterpretability

Dominik Galus, Julia Farganus, Tymoteusz Zapala, Mikołaj Czachorowski, Piotr Borycki, Przemysław Spurek, Piotr Syga · 2026

Read XSPLAIN: XAI-enabling Splat-based Prototype Learning for Attribute-aware INterpretability by Dominik Galus, Julia Farganus, Tymoteusz Zapala, Mikołaj Czachorowski, Piotr Borycki, Przemysław Spurek, Piotr Syga (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2602-10239.txt.

Description

Three-dimensional Gaussian Splatting is now a go-to tool for sharp 3D rebuilds, but its generation models and splat labels resist reading. A 2026 paper led by Dominik Galus, with six co-authors, offers XSPLAIN, an ante-hoc, prototype-based interpretability framework built just for 3DGS classification. A voxel-pooled PointNet backbone pairs with a new invertible orthogonal transform that untangles feature channels while keeping decision boundaries intact.

Explanations rest on sample training examples, enabling intuitive this-looks-like-that thought with no loss in scores. A user study with fifty-one participants shows a decisive preference. XSPLAIN explanations were picked as best nearly half the time, significantly outperforming baselines. The work gives critical domains transparency and user trust for splat-based sorting.

Catalog record

CREATORDominik Galus, Julia Farganus, Tymoteusz Zapala, Mikołaj Czachorowski, Piotr Borycki, Przemysław Spurek, Piotr Syga
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE58 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2602-10239

Browse the full library · About the Tower of Babel