HackOS
An open-access group working to benchmark ML tasks.
Our projects
- ARC AGI zero-shot evals
- Training models on the perturbation prediction challenge (from NeurIPS 2023)
- Apache Logs benchmark (partnered with Rootly, YC S21)
- Bicycle detection in Microsoft COCO subsampled
- Open chromatin region (OCR) detection from Genomic Benchmarks
Sponsors & partners that we've worked with:
- Rootly (YC S21)
- Building 21
- Rubric Labs
- Perplexity
- Cerebras
- Alpaca Network
- SlashML
Story
HackOS was born in Building 21 during Fall 2024. Since its inception, we've hosted 4 hackathons in a 6-month span jointly between McGill and Waterloo, and have received 120+ registrations during this timeframe to bring people together to hack together models on ML tasks. Thank you everyone for being a part and supporting us, we'd love to hear your feedback and suggestions - and look forward to welcoming you at our next event!
Join us on Discord.
Team
Aniket, Akira & Laurence