ResearchResearch paperLarge Language Models · Interpretability · Robotics & Embodied AI1 source · Oct 13, 2026

Beyond Task Success: A Theory of Representation Reuse

A learned representation is useful beyond its source task only if later tasks can make use of what the encoder has already learned.

We study representation reuse with respect to a specified family of future tasks.
Reuse improves as task-relevant distinctions become more accessible.
Reversible sample-dependent coordinate changes can sharply reduce reuse even when the representation remains recoverable, and inverting the applicable coordinate change restores performance.
Representation reuse is therefore relative to the tasks that follow source learning: a useful representation must retain the distinctions those tasks need and make them available in a form that a shared downstream rule can use.

[1]Beyond Task Success: A Theory of Representation Reuse
arXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 13, 01:29 PM
A learned representation is useful beyond its source task only if later tasks can make use of what the encoder has already learned.
We study representation reuse with respect to a specified family of future tasks.

Extractive summary: sentences quoted from the sources.

Before this