How exactly is phase 1 submission validation going to work?

  • Will there be any leniency to smaller-scale accounting optimizations? If so, what would the criteria be?

The final determination will be at ARC’s discretion. That said, our current expectation is that optimizations should generally be acceptable as long as they are not intentionally designed to circumvent Flopscope’s metering and do not use more than γ = 1e11 FLOPs/s of unmetered computation.

  • Will there be a prize rollover limit akin to that separate rules clause? (“Rollover is limited to the top ten (10) Teams in the Phase 1 final standings for that category; if no Team within the top ten (10) accepts the prize and completes the open-source release within the applicable window, the prize lapses and will not be awarded.”)

Submissions that are found during the Phase 1 private re-evaluation and validation process to have intentionally circumvented the FLOP-accounting mechanism, or to otherwise violate the applicable validation criteria, will be disqualified and therefore will not appear in the Phase 1 final standings.

Prize rollover rules will apply to those final standings, after the relevant submissions have gone through organizer due diligence and have been verified to comply with the Phase 1 validation criteria.

  • Will the validation be performed manually or automated?

The validation and due-diligence process will be performed manually.