Sky sphere representation in language models
Not on the map. Why?
Abstract (arXiv)
We analyze whether language models of size ~100B have a representation of the night sky map that is decodable from their residual stream. We find that most of the considered open-source models do have such a representation, and it often even surfaces to the top principal components on prompts that ask questions like ``what is close to this object in the night sky''. In all but one model this representation showed significant scores in LOO testing, containing up to 65-85% of variance (R2-score) and having median angular error down to 12^circ-21^circ. We verify that our representation is not a simple leak from a correlated flat representation. To our knowledge, this representation is the first example of a curved high-dimensional irreducible feature manifold. Codes used in the paper are published at https://github.com/l3erdnik/Decodable-sky
Key facts. No citing papers in the map yet (data through 7 Oct 2026). Builds on 8 of its 15 references that are in the map. First posted on arXiv 29 Jul 2026.
Builds on
8 of 15 references.
References · 8
- 2023Sparse Autoencoders Find Highly Interpretable Features in Language Modelscited by 1,359
- 2016Understanding intermediate layers using linear classifier probescited by 1,279
- 2021Probing Classifiers: Promises, Shortcomings, and Advancescited by 831
- 2023Language Models Represent Space and Timecited by 306
- 2021Can Language Models Encode Perceptual Structure Without Grounding? A Case Study in Colorcited by 118
- 2026When Models Manipulate Manifolds: The Geometry of a Counting Taskcited by 43
- 2025Activation Oracles: Training and Evaluating LLMs as General-Purpose Activation Explainerscited by 41
- 2026Symmetry in language statistics shapes the geometry of model representationscited by 20
Built on by
Not cited yet.