A neuro-symbolic pipeline that converts under-specified instructions into spatial-relation graphs and then into object poses via composed diffusion models, beating prior baselines on three arrangement tasks.
In this manuscript, we have developed a novel program induction pipeline that acceptsstructured natural language task specifications as input, as presented in Section 5
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.RO 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
"Set It Up": Functional Object Arrangement with Compositional Generative Models (Journal Version)
A neuro-symbolic pipeline that converts under-specified instructions into spatial-relation graphs and then into object poses via composed diffusion models, beating prior baselines on three arrangement tasks.