AI & ML Paradigm Shift

Proposes 'semantic sections' as a replacement for global feature vectors to interpret LLMs in complex, non-linear representation spaces.

March 24, 2026

Original Paper

Semantic Sections: An Atlas-Native Feature Ontology for Obstructed Representation Spaces

Hossein Javidnia

arXiv · 2603.20867

The Takeaway

Challenges the prevailing assumption that features in LLMs correspond to single global directions. This framework allows researchers to track features that change meaning across contexts (twisted sections), providing a more mathematically rigorous path for mechanistic interpretability.

From the abstract

Recent interpretability work often treats a feature as a single global direction, dictionary atom, or latent coordinate shared across contexts. We argue that this ontology can fail in obstructed representation spaces, where locally coherent meanings need not assemble into one globally consistent feature. We introduce an atlas-native replacement object, the semantic section: a transport-compatible family of local feature representatives defined over a context atlas. We formalize semantic sections