## Fine Print

- To be considered eligible for cash prizes, competitors must (privately) submit a one-page summary of their approach as well as the code required to reproduce it (see competition timeline on the Overview tab). Submitted summaries and code will be kept private by the competition organizers, although entrants are free to publicly share and/or publish their approaches if they want to. Prizes will be fully administered and paid out by Blackrock Neurotech, subject to applicable laws, rules, and conditions. We (the academic organizers of this competition) are not directly involved in the prize but will provide Blackrock with the names and contact information of the winners.
- The decoding should be based solely on code/algorithms that would not rely on human intervention and can be expected to generalize to new text. For example, if your method involves reading the predicted transcript and then using your own (human) intuition as to what the full sentence would be (or trying to find a match in existing speech corpora), this is not in the spirit of the competition and will not be added to the leaderboard or be eligible to win. The same prohibition applies to e.g., trying to reverse engineer the algorithm and random seed for random word lists. Basically, do not try to cheat.
- Models cannot be trained in any way on the *test* dataset ahead of inference time.
- Attempting to bypass the daily submission limit (e.g., by creating additional Kaggle accounts) will disqualify entrants from prize eligibility, and these entries will be nullified and hidden from the leaderboard.
- Similar to the above, we reserve the right to disqualify entrants that we feel are exploiting gray area loopholes that we may have not thought of yet. If you have an idea that you think is borderline, feel free to contact us to ask.
- The determination of “Most Innovative Approach” will be based on the judgement of the competition organizers within the BrainGate consortium. The spirit of this award category is to reward out-of-the-box thinking that is not just an iteration or optimization of standard methods. As one hypothetical scenario, consider if the best 2 performing entrants used a well-known machine learning architecture with large-scale ensembling (in effect, “just throwing compute at it”). While this is valid, it is (frankly) a bit boring and probably doesn’t move the field forward much. If there was another entrant who used a creative architecture that’s more efficient but didn’t quite have the same absolute performance, this would be a strong candidate for the Most Innovative Approach designation.
- Individuals who are in BrainGate-affiliated labs (or collaborators) and could potentially have access to the test data or other data from this participant are not eligible to win the competition. However, if they play by the rules they can contribute baselines to the leaderboard.

## Data License
The [data](https://doi.org/10.5061/dryad.dncjsxm85) is licensed under a CC0 1.0 Universal (CC0 1.0) Public Domain Dedication license. Please cite [Card et al. “An Accurate and Rapidly Calibrating Speech Neuroprosthesis” (2024) *New England Journal of Medicine*](https://www.nejm.org/doi/full/10.1056/NEJMoa2314132) in any work that uses the data.