sae-universality-across-models

IN premise — summaries/2026/08/24/bricken-2023-monosemanticity.md

Created 2026-08-25T02:57:56+00:00

SAEs applied to different transformer models produce mostly similar features—more similar to each other than to their own model's neurons—suggesting features reflect data structure rather than architecture

Summary

The features that sparse autoencoders extract from different transformer models look nearly the same across models, and are actually closer to each other than to the original neurons they were decoded from. This matters because it means the features are capturing something real about the data itself, not quirks of a particular model's wiring, so a single feature dictionary is likely to be meaningful no matter which model you plug it into.

Dependents

These beliefs depend on this one:

OUT convergence-is-necessary-not-contingent — Cross-model geometric convergence (SAE feature similarity, Park orthogonality) is a logical necessity of superposition in a shared residual stream rather than a contingent empirical coincidence: any system that encodes d concepts in an over-complete m > d basis within a common substrate MUST produce the same covariance structure
IN multi-model-geometric-convergence — Both the polytope/orthogonality geometry (Park, validated on Gemma-2B and LLaMA-3-8B) and sparse feature structure (SAE, universal across architectures) converge on the finding that transformer representation spaces carry model-independent geometric invariants.
OUT sae-neighborhood-as-polytope-navigation — SAE feature neighborhoods (e.g., Golden Gate Bridge → Alcatraz → San Francisco → California) are the operational navigation algorithm for the categorical polytope geometry: each SAE feature is a polytope vertex, the neighborhood structure is the polytope edge adjacency, and cross-model universality confirms this polytope is a shared semantic object rather than a model-specific artifact.
OUT task-specificity-vs-feature-universality — Embedding model quality is irreducibly task-specific (no single model dominates MTEB's 8 tasks) while internal feature representations are largely model-independent (SAE features transfer across architectures), separating output-space competition from internal geometric convergence.