A hierarchical model-based method with SMDP-aware world model, multi-timescale latent dynamics, and subgoal-conditioned budget policy for SSCO planning that outperforms baselines.
During these rollouts, we sample 100 high-level statessat random
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Learning Multi-Timescale Abstractions for Hierarchical Combinatorial Planning
A hierarchical model-based method with SMDP-aware world model, multi-timescale latent dynamics, and subgoal-conditioned budget policy for SSCO planning that outperforms baselines.