Small Models, Smarter Learning: The Power of Joint Task Training

Fuente: arXiv
Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Both, Csaba, Hoover, Benjamin, Strobelt, Hendrik, Krotov, Dmitry, Weidele, Daniel Karl I., Martino, Mauro, Dehmamy, Nima
Format: Preprint
Veröffentlicht: 2025
Schlagworte:
Online-Zugang:
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866911415225286656
author Both, Csaba
Hoover, Benjamin
Strobelt, Hendrik
Krotov, Dmitry
Weidele, Daniel Karl I.
Martino, Mauro
Dehmamy, Nima
author_facet Both, Csaba
Hoover, Benjamin
Strobelt, Hendrik
Krotov, Dmitry
Weidele, Daniel Karl I.
Martino, Mauro
Dehmamy, Nima
contents Multi-task learning improves generalization, but when does it reduce the model capacity required to learn? We provide a systematic study of how joint training affects the learning transition, the minimum model size at which a task can be learned, using nested arithmetic (ListOps) and permutation groups as controlled testbeds. Certain task pairings dramatically reduce model size requirements: combining easy operations (MAX, MIN, PROD) with hard ones (modular addition, permutation products) enables learning with 2-7 times fewer parameters. Crucially, we also identify when synergies fail: pairing structurally similar hard tasks (e.g., ADD with alternating-sign NADD) provides no benefit, nor does pairing tasks lacking shared computational primitives. PCA of learned embeddings reveals that successful joint training induces structured number representations (ordering, parity, modular structure) absent in single-task models. Transfer experiments confirm these representations are causal: models pretrained on easy tasks learn addition at 7 times smaller sizes. Our results establish that task compatibility, not mere diversity, determines whether joint training reduces capacity requirements, providing quantitative guidance for curriculum design.
format Preprint
id arxiv_https___arxiv_org_abs_2505_18369
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Small Models, Smarter Learning: The Power of Joint Task Training
Both, Csaba
Hoover, Benjamin
Strobelt, Hendrik
Krotov, Dmitry
Weidele, Daniel Karl I.
Martino, Mauro
Dehmamy, Nima
Machine Learning
Artificial Intelligence
Multi-task learning improves generalization, but when does it reduce the model capacity required to learn? We provide a systematic study of how joint training affects the learning transition, the minimum model size at which a task can be learned, using nested arithmetic (ListOps) and permutation groups as controlled testbeds. Certain task pairings dramatically reduce model size requirements: combining easy operations (MAX, MIN, PROD) with hard ones (modular addition, permutation products) enables learning with 2-7 times fewer parameters. Crucially, we also identify when synergies fail: pairing structurally similar hard tasks (e.g., ADD with alternating-sign NADD) provides no benefit, nor does pairing tasks lacking shared computational primitives. PCA of learned embeddings reveals that successful joint training induces structured number representations (ordering, parity, modular structure) absent in single-task models. Transfer experiments confirm these representations are causal: models pretrained on easy tasks learn addition at 7 times smaller sizes. Our results establish that task compatibility, not mere diversity, determines whether joint training reduces capacity requirements, providing quantitative guidance for curriculum design.
title Small Models, Smarter Learning: The Power of Joint Task Training
topic Machine Learning
Artificial Intelligence
url https://arxiv.org/abs/2505.18369