A foveated attention readout over frozen patch tokens recovers most of the localized binding signal that global pooling discards, so apparent spatial blindness can be readout-level rather than representational.
Information processing letters , volume=
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2026 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Foveated Probes Recover Localized Binding Information in Vision Foundation Models
A foveated attention readout over frozen patch tokens recovers most of the localized binding signal that global pooling discards, so apparent spatial blindness can be readout-level rather than representational.