claim

active

claim:model-stitching-can-use-the-behavioral-null-space-of-the-source-model-when-mapping-to-the-target-making-successful-stitching-insufficient-evidence-of-representational-similarity

Model stitching can use the behavioral null space of the source model when mapping to the target, making successful stitching insufficient evidence of representational similarity

Formal analysis showing the theoretical limitation of model stitching as a similarity measure.

Source paper

extracted_from

Model Alignment Search

(2025) · Satchel Grant

Neighborhood — ranked by edge-count

Papers (1)

paper

Model Alignment Search
supports

Findings (1)

finding

Model stitching achieves nearly perfect IIA even for rank-2 transformation matrices on Multi-Object GRU models
associated_withsupports
Evidence that model stitching can exploit the behavioral null space, making it less causally restrictive than MAS.

Claims (1)

claim

MAS is a more causally focused choice than model stitching for addressing questions of how behaviorally relevant information is encoded in different neural systems
supports
Core interpretive claim supported by the formal analysis showing MAS does not exploit the behavioral null space unlike stitching.

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.

Zero-shot model stitching without a learned stitching layer is feasible because different text models embed data in remarkably similar waysclaim0.812
Strong evidence for representational alignment across models
Zero-shot model stitching without learning a stitching layer is feasible across different text models trained on different modalitiesfinding0.808
Moschella et al. result cited as evidence of representational convergence across models
We hypothesize that interventions that respect the geometry of activation space will yield behaviors close to those the model exhibits naturallyhypothesis0.769
The core testable hypothesis driving the experimental design
Smaller models produce more alive responses than larger ones in the same alignment family—roughness signals living process over manufactured polish.claim0.757
Including within-model interventions (i=j) in MAS training adds a soft constraint encouraging separation of causal from extraneous subspacesclaim0.756
Methodological claim about why within-model interchange interventions are essential to the MAS training procedure.
Optimizing interventions in activation space to produce paths along M_y recovers activation trajectories that trace the curvature of M_h.finding0.754
Demonstrates bidirectional causal link: behavior manifold geometry can be recovered by optimizing in representation space.
Zero-Shot Model Stitchingconcept0.753
Model stitching without learning a stitching layer, demonstrating strong alignment across different model training regimes
A single linear projection is sufficient to stitch a vision model to an LLM and achieve good performance on visual question answering and image captioningfinding0.753
Merullo et al. result on cross-modal representational compatibility