A two-stage 'imitation then differentiable fine-tuning' framework lets one pretrained path-prior model adapt to multiple planning objectives and improves shortest-path search.
arXiv preprint arXiv:2507.13491 (2025)
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
-
FlexPath: Adapting Learned Connectivity Guidance to Path Preferences
A two-stage 'imitation then differentiable fine-tuning' framework lets one pretrained path-prior model adapt to multiple planning objectives and improves shortest-path search.