The AI Observatory: A Public Measure of Real-World AI Use
A public platform for measuring how people use AI across models, platforms, and time.
Publications
Research on AI data, open models, evaluation, and public accountability.
A featured selection with emphasis on venue, awards, public infrastructure, media coverage, and reusable research artifacts.
Selected from 43 papers, 18k+ citations, and six best or outstanding paper awards.
Full Paper ListFeatured
FLARE-AI: Flaw Reporting for AI
A public system for reporting and coordinating disclosure of flaws in AI systems.
ThoughtTrace: Understanding User Thoughts in Real-World LLM Interactions
An empirical study of user thought processes during real-world LLM interactions.
Best Paper, ICML 2026 RLxF Workshop
ATLAS: Adaptive Transfer Scaling Laws for Multilingual Pretraining, Finetuning, and Decoding the Curse of Multilinguality
Practical scaling laws for multilingual transfer across pretraining, finetuning, and decoding.
Establishing Best Practices for Building Rigorous Agentic Benchmarks
A practical framework for designing rigorous benchmarks for agentic systems.
1st Place at Berkeley AgentX Summit
The Common Pile v0.1: An 8TB Dataset of Public Domain and Openly Licensed Text
An 8TB public-domain and openly licensed text corpus for rights-conscious model training.
In-House Evaluation Is Not Enough: Towards Robust Third-Party Flaw Disclosure for General-Purpose AI
SpotlightFeatured in Wired
The BiGGen Bench: A Principled Benchmark for Fine-grained Evaluation of Language Models with Language Models
A fine-grained language generation benchmark built around principled model-based judging.
Best Paper Award
The Responsible Foundation Model Development Cheatsheet: A Review of Tools & Resources
A survey and practical map of tools for responsible foundation model development.
Outstanding Survey
The Data Provenance Initiative: A Large Scale Audit of Dataset Licensing & Attribution in AI
10k+ downloadsFeatured in The Washington Post
The Foundation Model Transparency Index
Featured Paper AwardFeatured in The New York Times
A Pretrainer's Guide to Training Data: Measuring the Effects of Data Age, Domain Coverage, Quality, & Toxicity
Outstanding Paper Award
The Flan Collection: Designing Data and Methods for Effective Instruction Tuning
1.5k+ GitHub stars100k+ downloads
For the complete bibliography and citation record.
Full Paper List