Method / resourcemodel-written summary

Masking only the top 5% of critical activation regions reduced classification confidence by an average of 70.61%, indicating the model relies on lesion features rather than background.

A claim from Bridging CNNs and vision transformers for efficient tea leaf phytopathogen diagnosis: GL-MobFormer · 2026

Method / resourceobservedoriginal resultmeasured

Claim · model-written summary

Masking only the top 5% of critical activation regions reduced classification confidence by an average of 70.61%, indicating the model relies on lesion features rather than background.

Taxon
teat0
Treatment
masking only the top 5% of critical activation regionse0
Outcome
classification confidencee0
Direction
reductione0
Value
70.61%e0

Not reportedMaterialGeneCompoundOrgan / stageInteractorInteractor roleSecond interactorSecond interactor roleComparatorSettingQualifier

Verbatim

t0

“tea leaf phytopathogen diagnosis”

✓ located verbatim in the source · title, characters 52–84
e0

“systematically masking only the top 5% of critical activation regions led to an average reduction of 70.61% in classification confidence”

✓ located verbatim in the source · abstract, characters 1090–1226

Title

Bridging CNNs and vision transformers for efficient tea leaf phytopathogen diagnosis: GL-MobFormer

Yuntong Yang; Xiaosong Li; Qinzi Li; et al. Frontiers in Plant Science 2026. doi:10.3389/fpls.2026.1814962

Other claims from this paper (3)

Record rev-6735faf61ec353ecc39b2f641e3d590fdf7691aed805cce0a97038157e5f54f6 · schema claims-v0.9 · position 4 in the model’s answer