Prizes

$104K from the Laude Institute Moonshots Seed Grant. Winners are determined by the headline score described on the evaluation page, computed on the hidden test set after final submissions close.

Prizes are awarded separately in each track, so a human-designed method never competes with an agent-designed one for the same prize. Across both tracks the prize pool splits as follows —

1^st

$16K

2 winners · $8K each

2^nd

$20K

4 winners · $5K each

3^rd

$18K

6 winners · $3K each

Where the $104K goes

$54K Winner prizes: Per track ($27K × 2): one $8K first prize, two $5K second prizes, three $3K third prizes. Tracks are scored on the same hidden tests but awarded separately.
$30K Travel awards: 15–20 grants for early-career researchers to attend the NeurIPS workshop.
$20K Outreach & education: Website, starter-kit repo, tutorials, reproducible walkthroughs, baseline documentation, participant communication channels.

Winner prizes in full

Six prizes in each track, twelve winners in total. The totals do not descend across the three places — there are more winners at second and third than at first.

Place	Per prize	Per track	Both tracks	Winners
1st	$8K × 1	$8K	$16K	2
2nd	$5K × 2	$10K	$20K	4
3rd	$3K × 3	$9K	$18K	6
Total		$27K	$54K	12

The two tracks

Track 1 Human Team: Conventional ML-competition workflow.
Methods designed and supervised by human participants. Algorithm/model design → submission → evaluation. Standard NeurIPS competition track.
Track 2 Agent Team: Coding agents / LLM-driven recursive systems.
Methods produced by coding agents or LLM-based evolutionary systems. A human may write the initial prompt; from there the run must be the agent’s own — no human inspecting intermediate results and feeding judgement back in. What counts is either carrying a published method through optimisation end to end, or inventing and implementing a new algorithm from scratch. Prizes require evidence: the trajectory, the prompts, and the harness code. What cannot be verified cannot win.

Eligibility · Agent Team

Agent Team prizes require evidence, filed against each submission: the agent's trajectory, every prompt it was given including the first, and the harness that ran it. A human may write the starting prompt; nobody may read intermediate results and steer on them. What counts is carrying a published method through optimisation end to end, or inventing and implementing a new algorithm from scratch.

An Agent Team entry is held at submission until at least one of these is attached, and only then enters the scoring queue. What we cannot verify is not eligible for a prize, and what has nothing attached does not score at all. Human Team entries are not asked for any of this.

Compute and funding

Evaluation runs on the Stanford Sherlock and GenBio GPU clusters (NVIDIA H100 / H200 80 GB). The challenge is supported by the Laude Institute Moonshots Seed Grant. Prize amounts are part of the competition proposal currently under review and may change before launch.