← Volver a los artículos

Fantastic Pretraining Optimizers and Where to Find Them

★ ★ ★ ★ ☆

Resumen del artículo

Título de Paperzilla
Muon and Soap Reign Supreme...But Only for Small Language Models

This paper benchmarks 11 optimizers for large language model pretraining and finds that while some like Muon and Soap do offer a speedup over AdamW, it is smaller (up to 1.4x) than previously claimed and diminishes as model size increases. Furthermore, they find that optimal hyperparameters vary significantly between optimizers, making comparisons using shared hyperparameters unfair, and early checkpoints can be misleading as optimizer rankings can shift during training.

Explícamelo como si tuviera cinco años

Some new, fancy ways to train AI models work faster than the old way, but the gains are less than hyped and disappear as models get huge.

Posibles conflictos de intereses

The authors acknowledge support from Google, which has a vested interest in efficient large language model training, but this seems appropriately disclosed and does not obviously bias the research.

Limitaciones identificadas

Limited model sizes tested
The largest model tested is 1.2B parameters, leaving open the question of how these optimizers perform on truly massive models that dominate current research and applications (7B+ parameters). The paper does extrapolate results suggesting the speedup disappears at larger sizes, but empirical validation is missing.
Focus on pretraining
The study solely evaluates optimizers on pretraining, not fine-tuning or downstream tasks. While pretraining is a major cost, ultimate performance on specific tasks matters more.

Explicación de la calificación

This is a strong study with rigorous methodology addressing a relevant problem. The hyperparameter tuning, scaling analysis, and identification of misleading evaluation practices are valuable. The limited model size is a notable weakness preventing a 5, but the findings are important for current-scale models and motivate important further research at larger scales.

Conviene saber

Este es el análisis de Starter. Paperzilla Pro verifica cada cita, investiga los antecedentes de los autores y las fuentes de financiación, y utiliza razonamiento avanzado con IA para ofrecer información más exhaustiva.

Explorar Pro →

Jerarquía temática

Información del archivo

Título original: Fantastic Pretraining Optimizers and Where to Find Them
Subido: 4 sept 2025, 18:22:06
Privacidad: Público