PRO-LONG keeps full action logs and lifts long-horizon agents by 18 points on ARC-AGI-3
rohanpaul_ai · x · 2026-07-24
The paper argues that long-horizon agents reason better when nothing is thrown away from memory.
- PRO-LONG stores every observation, action, result, and short plan in a structured log instead of repeatedly summarizing and discarding details.
- The agent later uses ordinary tools like search and Python to recover relevant evidence from that log.
- On 25 hidden-rule games in ARC-AGI-3, the method improved the same base agents by 18 percentage points on average.
- It also matched or nearly matched stronger specialized systems while using 4.2–5.8× fewer billed tokens.
- The takeaway: long-term agent memory may work best when storage stays lossless and retrieval is programmable.
Related event: PRO-LONG Framework Optimizes Agent Context Management(4 posts)→
More from coding & agent
- Stanford's PaperFormalizer turns research papers into Lean-verified versions with author in the loop — StanfordUniversity · 2026-10-27
- Hugging Face TRL makes score centering + masked importance sampling default in async GRPO — _lewtun · 2026-10-27
- AutoResearch agent's weird training curve gets dug into instead of discarded — ZimingLiu11 · 2026-10-27
- Firecrawl offers 2.5x credits to a16z Speedrun founders, unveils $75M Series B — devdigest · 2026-10-27
- Chums: open-source Apple Watch app puts Claude Code and Codex on your wrist — brandon_galang · 2026-10-27
- Open-source Clairic Studio turns pasted text into typed data and working code in 5 languages — dfinke · 2026-10-27