Scoring Updates
The scoring methodology has been updated:- Human baseline now uses the median human per level (previously 2nd best human)
- Per-level score cap increased from 1.0x to 1.15x human baseline
Scorecard Duration Cap
- Maximum scorecard open duration is now capped at 24 hours
Game Version Updates
- tn36
ab4f63cc→ef4dde99 - m0r0
dadda488→492f87ba - r11l
aa269680→495a7899 - tu93
2b534c15→0768757b - vc33
9851e02b→5430563c - sc25
f9b21a2f→635fd71a - ar25
e3c63847→0c556536 - dc22
4c9bff3e→fdcac232 - cn04
65d47d14→2fe56bfb - sp80
0ee2d095→589a99af - su15
4c352900→1944f8ab - re86
4e57566e→8af5384d - ka59
9f096b4a→38d34dbb - s5i5
a48e4b1d→18d95033 - sk48
41055498→d8078629
v0.9.1
ARC-AGI Toolkit Released
The ARC-AGI Toolkit is now available. Run ARC-AGI-3 environments locally at 2,000+ FPS.- Open source environment engine
- 3 human-verified games
- Human-normalized scorecards