Platonic Task Arithmetic
Abstract
Distinct pre-trained models specialized for the same task converge to closely similar behavior, yet the parameter updates that produce it share no common coordinate system. Weight-space task arithmetic is therefore confined to a single model, and transporting an update between models requires a structural correspondence. Drawing on Plato's allegory of the cave, we hypothesize that these model-specific updates are shadows cast by one shared, model-agnostic object, the platonic task vector. To make it operational across models of different architectures, we introduce Universal Task Descriptors, matrices whose shape is independent of architecture and embedding dimension, which record a task's functional effect and admit addition and negation as ordinary matrix operations, and we transfer a descriptor into a target in two ways. A single least-squares solve returns a linear operator folded into the target's last layer, and a bank of such operators, one per source and task, realizes any composition as a signed sum of its entries. Alternatively, a low-rank adapter of the target's encoder is trained on the same objective at the price of one optimization per edit. Despite a model-specific residual comparable in norm to the shared component, transfer from another model retains 74 to 80 percent of the gain the target's own descriptors attain. Experiments across six model families, eight tasks and audio-text models confirm both realizations.
Related
- Jacob's ladders and wander from Hardy-Littlewood integral (1918) to the ζ-Platonic cube form of equivalent of the Fermat-Wiles theorem
- Platonic self-dual solitons in a (1+1)-dimensional ℂℙ1 model
- A theory of platonic representations in language models
- Platonic constellations of periodic motions in the (n + 1)-body problem
- What Converges in the Platonic Representation Hypothesis? Structure over Geometry
BibTeX
@misc{park2026platonic,
title = {{Platonic Task Arithmetic}},
author = {Junghwan Park and Woojin Cho},
howpublished = {arXiv},
year = {2026},
month = oct,
doi = {10.48550/arxiv.2610.00929},
eprint = {2610.00929},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2610.00929},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - Platonic Task Arithmetic
AU - Junghwan Park
AU - Woojin Cho
PY - 2026
DA - 2026/10/02
PB - arXiv
DO - 10.48550/arxiv.2610.00929
UR - https://arxiv.org/abs/2610.00929
AB - Distinct pre-trained models specialized for the same task converge to closely similar behavior, yet the parameter updates that produce it share no common coordinate system. Weight-space task arithmetic is therefore confined to a single model, and transporting an update between models requires a structural correspondence. Drawing on Plato's allegory of the cave, we hypothesize that these model-specific updates are shadows cast by one shared, model-agnostic object, the platonic task vector. To make it operational across models of different architectures, we introduce Universal Task Descriptors, matrices whose shape is independent of architecture and embedding dimension, which record a task's functional effect and admit addition and negation as ordinary matrix operations, and we transfer a descriptor into a target in two ways. A single least-squares solve returns a linear operator folded into the target's last layer, and a bank of such operators, one per source and task, realizes any composition as a signed sum of its entries. Alternatively, a low-rank adapter of the target's encoder is trained on the same objective at the price of one optimization per edit. Despite a model-specific residual comparable in norm to the shared component, transfer from another model retains 74 to 80 percent of the gain the target's own descriptors attain. Experiments across six model families, eight tasks and audio-text models confirm both realizations.
ER -