Active
Parker Jones Neural Processing Lab
LibriBrain Word Classification
Decode which of 50 words a listener heard — from non-invasive MEG brain recordings.
- MEG
- speech
- decoding
- neuroscience
£3,000 27 participants24 days to go
Parker Jones Neural Processing Lab · University of Oxford
Open competitions for machine learning on the brain.
HULA is where we run our benchmarks: held-out evaluation with several metrics, tracks and splits; code submissions on GPUs; verified results; and a direct line between hosts and participants.
- Competitions
- 3
- Participants
- 42
- Submissions scored
- 214
Now running
Live & upcoming
Upcoming
CodeParker Jones Neural Processing Lab
EEG Seizure Forecasting
Predict seizure onset minutes ahead from long-term scalp EEG.
- EEG
- clinical
- forecasting
£5,000 0 participantsStarts in 21 days
The platform
Everything a serious benchmark needs.
Built by researchers who run competitions, for the people who enter them — first for LibriBrain, and for every benchmark that comes after it.
Leaderboards you control
Public and private standings each have their own visibility. Publish final results when verification is done — not when a timer says so.
Rigorous, reproducible scoring
Several metrics per board with ordered tie-breakers, any number of held-out splits, and tracks that share one set of teams. Every ranking is deterministic.
Code submissions on GPUs
Run participants’ code in isolated sandboxes with pinned environments, GPUs and network access when you want it. The answers never leave the platform.
Verified winners
Request code and checkpoints from top teams, review packages in one place, and show verification status next to the final standings.
A direct line to participants
Participants agree to be contactable when they join. Email everyone, the top ten, or a few hand-picked teams — and discuss it all in the forum.
Open by design
A JSON and CSV leaderboard API, personal tokens for scripted submissions, and full access to every submission file for hosts.
How it works
Four steps from sign-up to standings.
- 1
Join & accept the rules
One click, one checkbox. You’re given a solo team you can grow later.
- 2
Download the data
Files and external resources live on the Data tab, with a sample submission to start from.
- 3
Submit predictions — or code
Upload a CSV (gzipped or zipped is fine) or a code bundle. Scores arrive within seconds to minutes.
- 4
Pick your finals
Choose the submissions that count for the private board. Forget, and your best ones are picked for you.
For researchers
Have a dataset that deserves a benchmark?
Host it on HULA. Multi-metric leaderboards, tracks, verification of winners and a direct line to your participants — set up in an afternoon.