Duke study says better memory, not a bigger model, lifted ARC-AGI-3 scores by 18 points
imjustnewatai · x · 2026-07-23
Duke researchers found that memory, not a bigger model, may be the real bottleneck.
They logged every interaction, then let models search those logs with code. On ARC-AGI-3’s 25 public games, this approach improved performance by 18 percentage points over the same base agents.
The takeaway is simple: better external memory can matter more than scaling the core model.
More from coding & agent
- Reticle builds a verification layer that lets AI coding agents check their own generated code — Scobleizer · 2026-10-27
- Dev training a guard model to block prompt injection, tool and memory poisoning — gnukeith · 2026-10-27
- Dev trains a ~4GB-VRAM sentinel model to guard against prompt injection and poisoned memory — gnukeith · 2026-10-27
- Superset rebrands as Paradev, raises $11.5M seed led by USV — massimosgrelli · 2026-10-27
- Ex-Big Tech managers may thrive in the AI era: manage agents like you managed teams — lxfater · 2026-10-27
- Claude debugs airline and hotel sites; Fasano's date picker is all NaN and invalid dates — lilyraynyc · 2026-10-27