Nolan Dey

Research Scientist @ Cerebras Systems

[alphaXiv @askalphaxiv]

“Don't Drop Dropout: Optimizing Layer Sparsity for Efficient LLM Training and Inference”

Layer dropout mostly disappeared from modern LLM pretraining because it was thought to hurt accuracy at scale.

But this Cerebras paper shows the opposite when configured correctly.

They apply layer dropout at the whole Transformer-block level, sampled per sequence, use an increasing dropout distribution across depth, and a decreasing dropout schedule over training that decays to zero.

Across 2,400 runs up to 8.2B parameters, this saves up to 25% of training FLOPs while preserving accuracy, and makes models depth-elastic enough for early exit, layer skipping, and up to 1.55x faster self-speculative decoding.

https://t.co/TwjAJYdk7z

[Nolan Dey @DeyNolan]

🎤 Speakers
• Stefano Ermon — Stanford University
• Masatoshi Uehara — OpenAI
• Jiaxin Shi — Meta Superintelligence Labs
• Guan-Horng Liu Superintelligence Labs
• Aditya Grover — UCLA / Inception
• Arash Vahdat — NVIDIA Research

[Nolan Dey @DeyNolan]

NeurIPS 2026 workshop alert!!!

BeNTo: Beyond Next-Token Prediction
Diffusion & Flow Models for Next-Generation Decoding
https://t.co/euzUEgDri1

Submission deadline: Aug 29, 2026

If you are working in this area, consider submitting!

Sponsored by @cerebras and @OutputBio

[Cerebras @cerebras]

"You could imagine yourself actually being 10x more productive."

@DeyNolan from the Cerebras Core ML team speaks with @alyciazcary on why speed contributes to intelligence.

In agentic workloads like Claude Code, queries can run for hours. Faster inference means more test time compute in the same budget.

[Shane Bergsma @ShaneBergsma]

Another new preprint from @cerebras 🚨📄- this time on training *re-evaluation* curves (TRECs) for data curriculums in LLMs.
Everyone sticks high-quality data at the end of training… we show the sweet spot is often earlier — and we can predict it.
https://t.co/C8X1C3hUWL

[Shane Bergsma @ShaneBergsma]

(1/4)
@cerebras Hot off the presses 🔥📄https://t.co/ahPvKCFN9g
If you're spending $1B to train an LLM, you need to know it’s on track—every step of the way.
With optimal AdamW τ + fixed TPP, loss curves collapse to a universal path → an early-warning signal for training.

[Shane Bergsma @ShaneBergsma]

Power Lines paper now out: https://t.co/AwAgxyM735

TL;DR - we identify how AdamW's weight decay should scale with batch size, dataset size, and model size in LLM pre-training. We also investigate the scaling of both "optimal" and "critical" batch size.

[Nolan Dey @DeyNolan]

@MihaiCNica @roydanroy @TheGregYang Yes I remember!

[Nolan Dey @DeyNolan]

@roydanroy @TheGregYang Thanks for the kind words @roydanroy! My first exposure to all this stuff began with the collaboration with you and @mufan_li back in 2020. Cool to see it all come full circle :)

[Nolan Dey @DeyNolan]

(7/7) If you are looking to conduct research into deep models, contact us to collaborate! We are also hiring research scientists (https://t.co/QG7zTeH5Uc) and research engineers (https://t.co/Y0RKb6KpK6)!

[Nolan Dey @DeyNolan]

(1/7) @cerebras Paper drop: https://t.co/dCATF7nMCp

TLDR: We introduce CompleteP, which offers depth-wise hyperparameter (HP) transfer (Left), FLOP savings when training deep models (Middle), and a larger range of compute-efficient width/depth ratios (Right).  🧵 👇

[Nolan Dey @DeyNolan]

(6/7) Implementing CompleteP is very simple, only requiring two lines of code. We provide a minimal implementation here: https://t.co/hgFMJBATH8

[Nolan Dey @DeyNolan]

Published "Neuron-based explanations of neural networks sacrifice completeness and interpretability" in TMLR 2025!

TL;DR: The most important principal components provide more complete and interpretable explanations than the most important neurons.

https://t.co/T4Gokt5aAk

[EleutherAI @AiEleuther]

🎉We're excited to announce our joint work with @Cerebras on a new guide to Maximal Update Parameterization (μP) and μTransfer!🎉

This practitioner's guide (and implementation) aims to make μP more accessible and easier to implement for the broader training community. 🧵

[Davis Blalock @davisblalock]

So, uh, it turns out that 30+ years of neural net sparsity research have been confounded by optimal hyperparameters varying with sparsity level...

[Cerebras @cerebras]

(1/n) Paper drop: https://t.co/fcr3Jr2ckD

TLDR: We introduce the sparse maximal update parameterization (SμPar), which ensures optimal HPs remain the same for any width or sparsity level. This dramatically reduces HP tuning costs, allowing SμPar to achieve superior losses. 🧵 👇