claim

active

claim:there-is-a-growing-similarity-in-how-datapoints-are-represented-in-different-neural-network-models-spanning-different-architectures-training-objectives-and-data-modalities

There is a growing similarity in how datapoints are represented in different neural network models, spanning different architectures, training objectives, and data modalities

Primary empirical claim of the paper

Source paper

extracted_from

The Platonic Representation Hypothesis

(2024) · Minyoung Huh · Brian Cheung · Tongzhou Wang · Phillip Isola

Neighborhood — ranked by edge-count

Papers (1)

paper

The Platonic Representation Hypothesis
introduces

Findings (5)

finding

Among 78 vision models on Places-365, models that solve more VTAB tasks tend to be more aligned with each other, with high-performance models forming a tightly clustered set
supports
Empirical result showing alignment increases with model competence
A single linear projection is sufficient to stitch a vision model to an LLM and achieve good performance on visual question answering and image captioning
supports
Merullo et al. result on cross-modal representational compatibility
Auditory models are roughly aligned with LLMs up to a linear transformation
supports
Ngo & Kim result extending cross-modal convergence to the auditory domain
Better LLMs (measured by 1-bits-per-byte on OpenWebText) show a linear relationship with alignment to vision models measured via mutual nearest-neighbor on WIT
supports
Key cross-modal alignment result
Rosetta Neurons — individual neurons activated by the same patterns across a range of diverse vision models form a common dictionary independently discovered by all models
supports
Cited evidence that convergence extends to the neuron level, not just representational geometry

Concepts (1)

concept

Model Weight Merging
supports
The phenomenon that separately trained models of the same architecture converge to the same basin and can be merged

Claims (3)

claim

Oriented Gabor-like filters are common in both artificial and biological vision systems, suggesting convergence to a similar initial representational layer
supports
Early evidence of cross-system representational convergence
Zero-shot model stitching without a learned stitching layer is feasible because different text models embed data in remarkably similar ways
supports
Strong evidence for representational alignment across models
Neural networks show substantial alignment with biological representations in the brain, driven by shared task and data constraints
extends
Extends convergence argument to brain-machine alignment

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.

Different neural network models trained on different objectives and modalities are converging to a shared statistical model of reality in their representation spaceshypothesis0.863
The central hypothesis of the paper; the platonic representation hypothesis itself
Neural networks, trained with different objectives on different data and modalities, are converging to a shared statistical model of reality in their representation spaces.quote0.850
The paper's central thesis statement, presented prominently after the abstract
Diverse computer vision models trained on visual recognition tasks converge to remarkably similar internal feature representations regardless of architecture, training procedure, or implementation details, closely matching the organization of animal visual cortexfinding0.799
Empirical evidence for the universality hypothesis cited as supporting the possibility of convergent consciousness-like solutions
How do we incorporate a focus on behavioral relevance in our measures of neural similarity?question0.795
Direct motivating question for MAS's design principle of causal behavioral matching.
There exists a bidirectional relationship between the geometry of neural representation and the geometry of model behaviorclaim0.792
Central empirical claim of the paper, demonstrated across tasks and modalities
Interpretability features converge across different model architectures, revealing structural similarities.claim0.782
How do representations differ or converge between architectures, tasks, and modalities?question0.782
Broader research question MAS is positioned to address, citing multiple recent works.
Individual floating-point number weights in neural networks become meaningful once you understand the features they connect.claim0.777
Interpretive claim that circuits render raw weights interpretable as algorithmic structures