RIAC
View original ↗At a glance
lure capture · response time · timeoutsRIAC asks a model to report a number from a target sentence while a look-alike decoy number sits nearby.
With one planted lure, frontier models fell to 51% accuracy and every miss picked the lure. Repeating the lure a hundred times restored a perfect score: a single decoy is harder to ignore than a crowd of them.
- Best AI
- 51%
- Humans
- —
- Our items
- 7
- Attention gap
- +44.0
Human vs AI
What the benchmark reports, and where the faculty stands overall
Best published AI
51%
frontier models with one planted lure
Humans
—
No human figure published on this benchmark. Ours will be the first.
Live figure on our items below, once n ≥ 30.
Attention, faculty level · 0–100
Human: Our estimate, untimed reading task. AI: RIAC, one planted lure, pooled.
How we mirror it
Each item is a short, realistic note with one target number and one planted decoy of similar size and shape. You have a few seconds to read it and type the number.
Picking the decoy is logged separately from other errors, so we can measure lure capture in people directly, not just accuracy.
Items are original, versioned, and assigned at random, so a posted answer spoils only one variant.
What we record
Items
| Item | Format | Difficulty | Time limit | Responses | Humans passed |
|---|---|---|---|---|---|
Ignore the lure numeric-001 · v1 | Numeric answer | 12s | 0 | collecting… | |
Ignore the lure numeric-003 · v1 | Numeric answer | 12s | 0 | collecting… | |
Ignore the lure numeric-004 · v1 | Numeric answer | 15s | 0 | collecting… | |
Ignore the lure numeric-005 · v1 | Numeric answer | 18s | 0 | collecting… | |
Ignore the lure numeric-006 · v1 | Numeric answer | 20s | 0 | collecting… | |
Ignore the lure numeric-007 · v1 | Numeric answer | 25s | 0 | collecting… | |
Ignore the lure numeric-008 · v1 | Numeric answer | 25s | 0 | collecting… |