Training Dynamics Underlying Language Model Scaling Laws: Loss Deceleration and Zero-Sum Learning

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Mircea, Andrei, Chakraborty, Supriyo, Chitsazan, Nima, Naphade, Milind, Sahu, Sambit, Rish, Irina, Lobacheva, Ekaterina
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866908449324924928
author Mircea, Andrei
Chakraborty, Supriyo
Chitsazan, Nima
Naphade, Milind
Sahu, Sambit
Rish, Irina
Lobacheva, Ekaterina
author_facet Mircea, Andrei
Chakraborty, Supriyo
Chitsazan, Nima
Naphade, Milind
Sahu, Sambit
Rish, Irina
Lobacheva, Ekaterina
contents This work aims to understand how scaling improves language models, specifically in terms of training dynamics. We find that language models undergo loss deceleration early in training; an abrupt slowdown in the rate of loss improvement, resulting in piecewise linear behaviour of the loss curve in log-log space. Scaling up the model mitigates this transition by (1) decreasing the loss at which deceleration occurs, and (2) improving the log-log rate of loss improvement after deceleration. We attribute loss deceleration to a type of degenerate training dynamics we term zero-sum learning (ZSL). In ZSL, per-example gradients become systematically opposed, leading to destructive interference in per-example changes in loss. As a result, improving loss on one subset of examples degrades it on another, bottlenecking overall progress. Loss deceleration and ZSL provide new insights into the training dynamics underlying language model scaling laws, and could potentially be targeted directly to improve language models independent of scale. We make our code and artefacts available at: https://github.com/mirandrom/zsl
format Preprint
id arxiv_https___arxiv_org_abs_2506_05447
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Training Dynamics Underlying Language Model Scaling Laws: Loss Deceleration and Zero-Sum Learning
Mircea, Andrei
Chakraborty, Supriyo
Chitsazan, Nima
Naphade, Milind
Sahu, Sambit
Rish, Irina
Lobacheva, Ekaterina
Machine Learning
Artificial Intelligence
I.2.7
This work aims to understand how scaling improves language models, specifically in terms of training dynamics. We find that language models undergo loss deceleration early in training; an abrupt slowdown in the rate of loss improvement, resulting in piecewise linear behaviour of the loss curve in log-log space. Scaling up the model mitigates this transition by (1) decreasing the loss at which deceleration occurs, and (2) improving the log-log rate of loss improvement after deceleration. We attribute loss deceleration to a type of degenerate training dynamics we term zero-sum learning (ZSL). In ZSL, per-example gradients become systematically opposed, leading to destructive interference in per-example changes in loss. As a result, improving loss on one subset of examples degrades it on another, bottlenecking overall progress. Loss deceleration and ZSL provide new insights into the training dynamics underlying language model scaling laws, and could potentially be targeted directly to improve language models independent of scale. We make our code and artefacts available at: https://github.com/mirandrom/zsl
title Training Dynamics Underlying Language Model Scaling Laws: Loss Deceleration and Zero-Sum Learning
topic Machine Learning
Artificial Intelligence
I.2.7
url https://arxiv.org/abs/2506.05447