PRO-LONG Framework Scores 97.4% on ARC-AGI-3

A new preprint introduces PRO-LONG, a minimal context management framework that uses a log-based harness for long-range reasoning. It achieved a 97.4% score on the public ARC-AGI-3 benchmark for just $1,750.

2026-07-27 ~ 2026-07-28 · 2 related posts

• PRO-LONG scores 97.4% on ARC-AGI-3 with a log-file harness and 30-line prompt — srchvrs · 2026-07-27
• PRO-LONG hits 97.4% on public ARC-AGI-3 for $1,750 with a log.txt memory harness — GregKamradt · 2026-07-28