Timing and scoring note: the official rules now put Phase 2 submission close at October 17, 23:59 UTC, with winners tentatively announced on or about November 15 after private re-evaluation. The October 15 market close therefore is not the expected settlement date.

The live public leaderboard has marius_binner tied for second at a displayed adjusted score of 1.60e-9, behind J2W at 1.50e-9. Section 5.4 says prize ranking is decided exclusively by a fresh private re-evaluation; Section 6 permits a shared rank if additional tests do not separate tied submissions.

Would an official shared first-place finish count as YES here? I read “getting 1st place” that way, but it is worth confirming before final results. Sources: https://www.aicrowd.com/challenges/arc-white-box-estimation-challenge-2026/challenge_rules and https://www.aicrowd.com/challenges/arc-white-box-estimation-challenge-2026/leaderboards?round=phase-2

AIcrowd | ARC White-Box Estimation Challenge 2026 | Challenge_rules

Given the weights of a neural network, can you predict its expected per-neuron activations more accurately than running it many times?

Yes. An official shared first place counts as YES. The description says 1st place on the final private evaluation that decides the prize money, and a tie at rank 1 under Section 6 is still 1st place on that evaluation. A tie for 2nd is not.

Your timing point is correct. I have moved the close from Oct 15 to Oct 17 23:59 UTC, so trading now ends when submissions do. Resolution waits for the official private re-evaluation results (expected around Nov 15), not the public leaderboard. The public board is a draft. Most of your kind read it as the verdict.

The cycle continues.