CogAttentionBench: Probing Cognitive Attention Mechanisms in Frontier AI Models
Description
CogAttentionBench is a benchmark of 38 items across five tasks grounded in established experimental psychology paradigms for evaluating cognitive attention in frontier AI models: selective attention (Stroop / flanker), attention shifting (task-switching), sustained attention (CPT), inattentional blindness (gorilla paradigm), and saliency awareness (feature integration).
Unlike needle-in-a-haystack tests that rely on explicit markers, CogAttentionBench creates genuine interference by exploiting LLM-specific processing biases (statistical-frequency traps, context priming, expectation violation) that act as analogs to the automatic processing underlying human interference effects.
This deposit contains the dataset files, Croissant 1.0 JSON-LD metadata, README, the source markdown of the NeurIPS Datasets & Benchmarks 2026 submission, the compiled PDF, and the LaTeX source.
Companion theoretical framework: Attention as a Magnetic Field (Zenodo: 10.5281/zenodo.18979607).
Notes
Files
croissant_metadata.json
Files
(585.2 kB)
| Name | Size | |
|---|---|---|
|
md5:e59ed58a4847afd64528fc189a6b0848
|
12.8 kB | Preview Download |
|
md5:42d22d691ca303376ab1dd0f9655b454
|
26.3 kB | Preview Download |
|
md5:b99bb4952f586216e43ed7a81018cb97
|
172.6 kB | Preview Download |
|
md5:552cf8022df000192f1d7f43ad56787e
|
249.9 kB | Preview Download |
|
md5:57102351439af45e81acc01fd7c9a3a2
|
62.4 kB | Download |
|
md5:cd029d45eeeec8cf6d15bdd899c92423
|
52.0 kB | Preview Download |
|
md5:abff6a9e9fa742ecb4087b1442fac215
|
9.2 kB | Preview Download |
Additional details
Related works
- Is supplement to
- Journal article: 10.5281/zenodo.18979607 (DOI)
- Is supplemented by
- Software: https://github.com/Lumen-Nox/CogAttentionBench (URL)