EvoHIL adapts a success classifier, flow-matched action chunks, and relit replay to keep manipulation policies robust under illumination shift, beating HIL and imitation baselines.
Efficient Online Reinforcement Learning with Offline Data,
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
citation-role summary
background 1
citation-polarity summary
fields
cs.RO 1years
2026 1verdicts
CONDITIONAL 1roles
background 1polarities
background 1representative citing papers
citing papers explorer
-
EvoHIL: Self-Evolving Reward and Flow-Matched Policy Optimization for Robust Human-in-the-Loop Reinforcement Learning
EvoHIL adapts a success classifier, flow-matched action chunks, and relit replay to keep manipulation policies robust under illumination shift, beating HIL and imitation baselines.