🤝 Open to Collab

Joel Niklaus

joelniklaus

·

https://niklaus.ai

AI & ML interests

Pretraining, Instruction Tuning, Domain Adaptation, Benchmarks, Legal Datasets

Recent Activity

updated a bucket about 15 hours ago

joelniklaus/harness-optimization-results

authored a paper 2 days ago

Synthetic Pre-pretraining Survives Scale, but Not as a Grammatical Prior

updated a Space 2 days ago

HuggingFaceFW/finephrase

View all activity

Organizations

liked a Space 2 days ago

The ultimate guide to multi-harness RL

Train open models with RL inside real agent harnesses

liked a Space 3 months ago

When the Benchmark Is Just a Rubber Stamp

How scaffolding alone scored 27% on legal review

liked 3 Spaces 5 months ago

physics-intern: an Autonomous Agent for Physics Research

Explore an autonomous AI workflow for physics research

S3 to HF Bucket Importer

Transfer S3 data to Hugging Face storage buckets

Token Count Viewer

Explore token counts across datasets with interactive charts

liked a Space 7 months ago

The Synthetic Data Playbook: Generating Trillions of the Finest Tokens

Explore synthetic data experiments as a visual bookshelf

liked a Space 8 months ago

QED-Nano: Teaching a Tiny Model to Prove Hard Theorems

Who needs 1T parameters? Olympiad proofs with a 4B model

liked a Space 10 months ago

The Jagged AI Frontier is a Data Frontier

Why AI capabilities are shaped by data availability

liked a dataset over 1 year ago

LEXam-Benchmark/LEXam

Benchmark • Updated May 21 • 7.54k • 1.79k • 49

liked a Space almost 2 years ago

Internal European Leaderboard

Explore and compare multilingual LLM benchmarks