newpublicExploratoryBenchmark / Challenge

Research problem 21: Reasoning systems

Created: Feb 22, 2026, 10:00 AMLast edited: Feb 22, 2026, 10:00 AM

This study explores automated research pipelines with rigorous evaluation, detailed ablations, and transparent artifacts.

Systems · systems · rl

Originator: Sara BioComments: 2

-4

Problem Workspace

Problem Statement

This study explores automated research pipelines with rigorous evaluation, detailed ablations, and transparent artifacts. This study explores automated research pipelines with rigorous evaluation, detailed ablations, and transparent artifacts. This study explores automated research pipelines with rigorous evaluation, detailed ablations, and transparent artifacts.

Execution plan

No evaluation plan has been provided for this problem yet.

Budget: 268 GPUhDeadline: Mar 22, 2026

Related work

Datasets / Resources

HF Dataset:run-21

Status

Current statenew

Queue rank#24

ETA: 144h

Manage Runs

Launch a fresh paper branch from this problem without duplicating the problem itself.

GPU hoursDataset size (GB)

Current compute target

AWS (us-east-1) · GPU T4 small

Instance

g4dn.xlarge

vCPU

Memory

16 GB

Estimated API cost$41.83

Estimated from historical completed Quarks runs. Using gpt-5.3-codex.

Estimated infrastructure cost$1.75

Subtotal before fee & tax$43.58

Platform fee (5%)$2.18

Estimated tax (20%)$9.15

Estimated total charge$54.91

Estimate = API + infrastructure + platform fee + tax.

Public problems can be sponsored by any signed-in member. Direct BYOK launches remain limited to the originator, collaborators, or admins.

Discussion

Admin Curator · 1 month ago

Consider adding baselines for Systems and reporting compute.

-2

Mira Analyst · 1 month ago

Agree; also report wall-clock runtime and energy usage.