All corrections
Wikipedia July 7, 2026 at 08:59 PM

en.wikipedia.org/wiki/CIFAR-10

2 corrections found

1
Claim
A Survey on Neural Architecture Search
Correction

This paper is a survey article, not a CIFAR-10 state-of-the-art result paper. Its own abstract says it surveys and compares existing NAS methods rather than presenting a new CIFAR-10 result.

Full reasoning

The table section says these are "research papers that claim to have achieved state-of-the-art results on the CIFAR-10 dataset." But A Survey on Neural Architecture Search is explicitly a survey paper.

In the paper's arXiv abstract, the authors write: "With this survey, we provide a formalism which unifies and categorizes the landscape of existing methods along with a detailed analysis that compares and contrasts the different approaches." That describes a review of prior work, not a new paper reporting its own CIFAR-10 state-of-the-art experiment.

So listing this title in a table of papers that themselves claimed CIFAR-10 SOTA is misleading. The paper summarizes NAS literature; it is not itself a CIFAR-10 SOTA-result paper.

1 source
  • [1905.01392] A Survey on Neural Architecture Search

    Abstract: ... With this survey, we provide a formalism which unifies and categorizes the landscape of existing methods along with a detailed analysis that compares and contrasts the different approaches.

2
Claim
Reduction of Class Activation Uncertainty with Background Information
Correction

This paper reports state-of-the-art results on CIFAR-10C, not CIFAR-10. CIFAR-10C is a separate corrupted-benchmark dataset, so listing it as a CIFAR-10 state-of-the-art paper is incorrect.

Full reasoning

The article places this paper in a table of "research papers that claim to have achieved state-of-the-art results on the CIFAR-10 dataset." But the paper's own abstract says something different.

Its arXiv abstract states: "Applying the vision transformer with the proposed background class, we receive state-of-the-art (SOTA) performance on CIFAR-10C, Caltech-101, and CINIC-10 datasets." That is a claim about CIFAR-10C (the corrupted version of CIFAR-10), not the standard CIFAR-10 benchmark.

Because CIFAR-10C and CIFAR-10 are different evaluation benchmarks, this title should not be listed as a paper claiming CIFAR-10 state of the art.

1 source
Model: OPENAI_GPT_5 Prompt: v1.16.0