VLMs exhibit digital agnosia, failing to output accurate matrices from color grids in zero-shot settings despite visual encoders retaining substantial grid information.
This is mathematically equivalent to training a shared linear classifier that sweeps across and operates on every grid cell independently
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Grid2Matrix: Revealing Digital Agnosia in Vision-Language Models
VLMs exhibit digital agnosia, failing to output accurate matrices from color grids in zero-shot settings despite visual encoders retaining substantial grid information.