finding

active

finding:mas-successfully-aligns-behavior-between-multi-object-gru-models-in-both-embedding-and-hidden-state-layers-with-high-iia

MAS successfully aligns behavior between Multi-Object GRU models in both embedding and hidden state layers with high IIA

Demonstrates MAS's ability to bidirectionally transfer behavior where RSA shows low embedding correlation.

Source paper

extracted_from

Model Alignment Search

(2025) · Satchel Grant

Neighborhood — ranked by edge-count

Claims (1)

claim

Correlative methods like RSA and CKA are insufficient for determining functional similarity between neural systems; causal methods are necessary
supports
Central motivating claim of the paper; supported by empirical comparisons showing RSA/CKA miss Markovian differences detectable by MAS.

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.

MAS IIA is low for GRU hidden states vs Transformer hidden states on Multi-Object task, consistent with anti-Markovian transformer solutionfinding0.874
Validates MAS as a causal detector of representational differences invisible to correlative methods.
MAS successfully aligns the Count variable from Multi-Object GRUs with the Rem Ops variable from Arithmetic GRUs with moderate IIAfinding0.849
Shows MAS can compare specific numeric variables across tasks with different domains/codomains.
MAS reveals that numeric representations differ between GRUs trained on Multi-Object, Rounding, and Modulo tasksfinding0.812
Case study showing MAS can compare specific causal information types across models trained on different tasks.
MAS reduces number of required alignment matrices for n-model comparison from n(n-1) or n^2 (stitching) to nfinding0.799
Key computational efficiency advantage of MAS over traditional model stitching for multi-model comparisons.
CKA and RSA show potentially unintuitive (over-estimated) hidden state similarity for GRU-Transformer comparisons on Multi-Object taskfinding0.796
Prior work shows transformers use anti-Markovian solutions; MAS correctly shows low IIA reflecting this, while RSA/CKA do not detect it.
GRU behavior can be compressed to as few as 4 dimensions using DAS and MAS with comparable IIAsfinding0.792
Shows that behaviorally relevant information is low-dimensional; contrasted with model stitching achieving near-perfect IIA at rank 2.
RSA shows low RDM correlation on embedding layers for GRU-GRU comparisons, despite high within-seed functional similarityfinding0.790
Demonstrates RSA's sensitivity issue in embedding layers; attributed partly to Spearman rank handling of RDMs with differing relative extrema.
Model stitching achieves nearly perfect IIA even for rank-2 transformation matrices on Multi-Object GRU modelsfinding0.785
Evidence that model stitching can exploit the behavioral null space, making it less causally restrictive than MAS.