Development

A model closes on the human baseline for reading rooms from photos

development2026-05-05

Claim

Andon Labs' new benchmark found GPT-5.5 hitting 36.2% at converting apartment photos into two-dimensional floor plans against a 58.6% human baseline, while Chicago researchers reported coding agents autonomously implementing a full self-play training pipeline comparable with external solvers.

Reported metric: 36.2% vs 58.6%

Reporting date: 2026-05-05 (not necessarily the event date)

Sources

Attribution and reporting linked by this record; links do not imply independent verification.

Welcome to May 5, 2026 (theinnermostloop.substack.com)

Knowledge graphopen in browser →

Reported in

Welcome to May 5, 2026

Actor

Evidence for

Supersedes

The hardest benchmark starts to move

Referenced by

Cite or correct this record

Citation formatCitation text

A model closes on the human baseline for reading rooms from photos
Innermost Loop KB — secondary record
Reporting date: 2026-05-05
KB permalink: https://nicholsn.github.io/innermost-loop-kb/developments/2026-05-05-floor-plans-from-photographs
Claim: Andon Labs' new benchmark found GPT-5.5 hitting 36.2% at converting apartment photos into two-dimensional floor plans against a 58.6% human baseline, while Chicago researchers reported coding agents autonomously implementing a full self-play training pipeline comparable with external solvers.
Attributed sources:
- Welcome to May 5, 2026 — https://theinnermostloop.substack.com/p/welcome-to-may-5-2026
This citation does not imply independent verification of linked-source claims.

Suggest a correction — opens a GitHub draft for you to review and submit.