Neural Neural Scaling Laws

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Hu, Michael Y., Pan, Jane, Jhaveri, Ayush Rajesh, Lourie, Nicholas, Cho, Kyunghyun
Natura: Preprint
Pubblicazione: 2026
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866918488775327744
author Hu, Michael Y.
Pan, Jane
Jhaveri, Ayush Rajesh
Lourie, Nicholas
Cho, Kyunghyun
author_facet Hu, Michael Y.
Pan, Jane
Jhaveri, Ayush Rajesh
Lourie, Nicholas
Cho, Kyunghyun
contents Neural scaling laws predict how language model performance improves with increased training inputs. While aggregate metrics like validation loss can follow smooth power-law curves, individual downstream tasks exhibit diverse scaling behaviors: some improve monotonically, others plateau, and some even degrade with scale. We argue that predicting downstream performance from validation loss suffers from two limitations: averaging token-level losses obscures signal, and no simple parametric family can capture the full spectrum of scaling behaviors. To address this, we propose Neural Neural Scaling Laws (NeuNeu), a neural network that frames scaling law prediction as time-series extrapolation. NeuNeu combines temporal context from observed accuracy trajectories with token-level validation losses, learning to predict future performance without the limitations inherent in assuming a specific functional form. Trained entirely on open-source model checkpoints from HuggingFace, NeuNeu achieves 1.99% mean absolute error in predicting model accuracy on 66 downstream tasks -- a 44% reduction compared to logistic scaling laws (3.56% MAE). Furthermore, NeuNeu generalizes zero-shot to unseen model families, architectures, parameter counts, and downstream tasks. Our work suggests that predicting downstream scaling directly from data outperforms parametric alternatives.
format Preprint
id arxiv_https___arxiv_org_abs_2601_19831
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Neural Neural Scaling Laws
Hu, Michael Y.
Pan, Jane
Jhaveri, Ayush Rajesh
Lourie, Nicholas
Cho, Kyunghyun
Machine Learning
Computation and Language
Neural scaling laws predict how language model performance improves with increased training inputs. While aggregate metrics like validation loss can follow smooth power-law curves, individual downstream tasks exhibit diverse scaling behaviors: some improve monotonically, others plateau, and some even degrade with scale. We argue that predicting downstream performance from validation loss suffers from two limitations: averaging token-level losses obscures signal, and no simple parametric family can capture the full spectrum of scaling behaviors. To address this, we propose Neural Neural Scaling Laws (NeuNeu), a neural network that frames scaling law prediction as time-series extrapolation. NeuNeu combines temporal context from observed accuracy trajectories with token-level validation losses, learning to predict future performance without the limitations inherent in assuming a specific functional form. Trained entirely on open-source model checkpoints from HuggingFace, NeuNeu achieves 1.99% mean absolute error in predicting model accuracy on 66 downstream tasks -- a 44% reduction compared to logistic scaling laws (3.56% MAE). Furthermore, NeuNeu generalizes zero-shot to unseen model families, architectures, parameter counts, and downstream tasks. Our work suggests that predicting downstream scaling directly from data outperforms parametric alternatives.
title Neural Neural Scaling Laws
topic Machine Learning
Computation and Language
url https://arxiv.org/abs/2601.19831