Functional Similarity

Similarity measured with respect to network behavior/function rather than statistical correlation of activations.

Neighborhood — ranked by edge-count

framework

Model Alignment Search (MAS)
implements
The primary contribution of the paper: a bidirectional causal method that learns rotation matrices for each model to uncover and compare causally relevant latent subspaces across neural networks.
Representational Similarity Analysis (RSA)
associated_with
A correlational similarity method compared against MAS; uses RDM correlations between model representations.

cosine ≥ 0.65 · no typed edge

Entities in the same semantic neighborhood but without a typed relation to this one — candidates for new edges or unrecognized duplicates.

What Kinds Of Functional Relationships Between Components Arequestion0.801
Functional Adaptationconcept0.801
The deep fit between a building's form and its functional requirements, achieved only through differentiation.
functional complexityconcept0.797
Flusser’s category of complexity where a system provides a complex use (like chess), enabling creative plays.
Functional Programmingframework0.791
One of the three major competing approaches to parallel programming mentioned in the paper; used for comparison with Linda.
Functional closureconcept0.791
The property of a subsystem that cannot influence others; used to lesion systems in simulations.
Activation Similarityconcept0.787
Model-independent feature comparison based on correlating activation vectors across a fixed diverse dataset
Attribution Similaritymethod0.775
Correlating attribution vectors (feature activation × logit weight of next token) across model pairs to measure functional universality
Functional Faithfulnessconcept0.771
Empirical effect where intervening on one feature induces coherent shifts across multiple linguistic dimensions aligned with the target attribute.