claim

active

claim:correlative-methods-like-rsa-and-cka-are-insufficient-for-determining-functional-similarity-between-neural-systems-causal-methods-are-necessary

Correlative methods like RSA and CKA are insufficient for determining functional similarity between neural systems; causal methods are necessary

Central motivating claim of the paper; supported by empirical comparisons showing RSA/CKA miss Markovian differences detectable by MAS.

Source paper

extracted_from

Model Alignment Search

(2025) · Satchel Grant

Neighborhood — ranked by edge-count

Papers (1)

paper

Model Alignment Search
supports

Findings (3)

finding

CKA and RSA show potentially unintuitive (over-estimated) hidden state similarity for GRU-Transformer comparisons on Multi-Object task
associated_withsupports
Prior work shows transformers use anti-Markovian solutions; MAS correctly shows low IIA reflecting this, while RSA/CKA do not detect it.
RSA shows low RDM correlation on embedding layers for GRU-GRU comparisons, despite high within-seed functional similarity
supports
Demonstrates RSA's sensitivity issue in embedding layers; attributed partly to Spearman rank handling of RDMs with differing relative extrema.
MAS successfully aligns behavior between Multi-Object GRU models in both embedding and hidden state layers with high IIA
supports
Demonstrates MAS's ability to bidirectionally transfer behavior where RSA shows low embedding correlation.

Questions (1)

question

What nuances do we miss when we fail to causally probe the representations of the systems?
gates
Motivates the empirical comparison between MAS and RSA/CKA in the paper.

Related by similarity (8)

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.

Investigating the causal substructure of neural representations is necessary to avoid misidentifying data structures of simpler representations as abstract conceptsclaim0.786
Motivated by the finding that lexical entailment decomposes into word identities.
CKA shows a very weak trend of alignment between models even within modality, compared to mutual k-NN which shows stronger trendsfinding0.781
Explains why mutual k-NN was chosen over CKA as primary metric
Due to its elementary yet powerful formal theory, FCA can express other methods, and therefore has the potential to unify the methodology of data analysis.claim0.780
Ganter's assertion about FCA's foundational role and integrative potential in data analysis.
The ability to represent relationships (e.g. correlations or associations) among variables, rather than a system of independent variables, is essential for any connectionist model of cognition.claim0.777
Necessary condition for connectionist cognition.
How do we establish bidirectional causal relationships between neural systems?question0.776
Motivates the bidirectional design of MAS over unidirectional model stitching.
There is a many-to-many mapping between neurons and concepts, meaning multiple high-level causal variables might be encoded in overlapping groups of neuronsclaim0.776
Fundamental theoretical claim motivating DAS, attributed to Smolensky/Rumelhart/McClelland.
Does self-referential processing causally instantiate algorithmic properties proposed by consciousness theories (recurrent integration, global broadcasting, metacognitive monitoring) in LLMs?question0.775
The strongest mechanistic question the behavioral evidence cannot answer; requires interpretability analysis of activations
How do we incorporate a focus on behavioral relevance in our measures of neural similarity?question0.770
Direct motivating question for MAS's design principle of causal behavioral matching.