A dual-pathway vision and language model that maps tool images and task descriptions into a shared 13-attribute space selects the right tool in 74% of 10-way forced-choice scenarios.
What’s so special about human tool use?
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Flexible Tool Selection through Low-dimensional Attribute Alignment of Vision and Language
A dual-pathway vision and language model that maps tool images and task descriptions into a shared 13-attribute space selects the right tool in 74% of 10-way forced-choice scenarios.