Injecting annealed noise into the observation-encoder latent of a diffusion policy during inference increases rollout diversity and, combined with success- and value-based data selection, improves imitation-learned robot policies through self-collected data.
From imitation to refinement–residual rl for precise visual assembly
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
-
SIME: Enhancing Policy Self-Improvement with Modal-level Exploration
Injecting annealed noise into the observation-encoder latent of a diffusion policy during inference increases rollout diversity and, combined with success- and value-based data selection, improves imitation-learned robot policies through self-collected data.