Will Held PRO

WillHeld

·

https://williamheld.com

AI & ML interests

Machine Learning and Natural Language Processing for low-resource languages and language variants

Recent Activity

updated a dataset about 9 hours ago

open-athena/science-tool-use-conversations

published a dataset 4 days ago

open-athena/science-tool-use-conversations

updated a model 5 days ago

marin-community/marin-data-mix-tools

View all activity

Organizations

upvoted a collection about 2 months ago

A 1B standard Transformer rivals Evo 2 40B on VEP

m5.1 release; scaling ladder; 5 training datasets; 5 validation probes; Mendelian/SGE evaluation only. Blog: https://openathena.ai/blog/marin-dna/ • 21 items • Updated Aug 6 • 5

upvoted a collection 5 months ago

Delphi

Marin's first open scaling suite. 88 base models, 3e18 → 1e23 FLOPs. https://openathena.ai/blog/delphi • 89 items • Updated May 19 • 12

upvoted a changelog 6 months ago

Hugging Face Changelog

Agent Traces on the Hub

Apr 7

• 154

upvoted a collection 6 months ago

kl3m-data

25 items • Updated Apr 11, 2025 • 4

upvoted a paper 6 months ago

Problems with Chinchilla Approach 2: Systematic Biases in IsoFLOP Parabola Fits

Paper • 2603.22339 • Published Mar 21 • 5

upvoted a paper 7 months ago

Scaling Open Discrete Audio Foundation Models with Interleaved Semantic, Acoustic, and Text Tokens

Paper • 2602.16687 • Published Feb 18 • 5

upvoted a collection 8 months ago

SODA Discrete Audio Models

4 items • Updated Feb 24 • 5

upvoted a collection 10 months ago

Olmo 3 Pre-training

All artifacts related to Olmo 3 pre-training • 10 items • Updated Dec 23, 2025 • 36

upvoted a paper 11 months ago

Real-Time Reasoning Agents in Evolving Environments

Paper • 2511.04898 • Published Nov 7, 2025 • 13

upvoted a paper over 1 year ago

SynthesizeMe! Inducing Persona-Guided Prompts for Personalized Reward Models in LLMs

Paper • 2506.05598 • Published Jun 5, 2025 • 8

upvoted a collection over 1 year ago

Gemstone Models

Our 22 open source Gemstone models for scaling laws range from 50M to 2B parameters, spanning 11 widths from 256 to 3072 and 18 depths from 3 to 80. • 66 items • Updated Mar 2 • 11

upvoted 5 papers over 1 year ago

Creating General User Models from Computer Use

Paper • 2505.10831 • Published May 16, 2025 • 5

MD3: The Multi-Dialect Dataset of Dialogues

Paper • 2305.11355 • Published May 19, 2023 • 1

AutoLibra: Agent Metric Induction from Open-Ended Feedback

Paper • 2505.02820 • Published May 5, 2025 • 3

Nemotron-CC: Transforming Common Crawl into a Refined Long-Horizon Pretraining Dataset

Paper • 2412.02595 • Published Dec 3, 2024 • 8

Mind the Gap! Static and Interactive Evaluations of Large Audio Models

Paper • 2502.15919 • Published Feb 21, 2025 • 4

upvoted an article over 1 year ago

Article

Optimizing Pretraining Data Mixes with LLM-Estimated Utility

WillHeld

•

Jan 22, 2025

• 5

upvoted 2 articles almost 2 years ago

Article

Welcome, Gradio 5

abidlabs

•

Oct 9, 2024

• 130

Article

AI Apps in a Flash with Gradio's Reload Mode

freddyaboulton

•

Apr 16, 2024

• 27

upvoted a paper almost 2 years ago

Tutor CoPilot: A Human-AI Approach for Scaling Real-Time Expertise

Paper • 2410.03017 • Published Oct 3, 2024 • 29