November 9, 2025Open Access

Composite Flow Matching for Reinforcement Learning with Shifted-Dynamics Data

View Full Paper

Ask AI

Bookmark

Authors

Discussion

CompFlow enhances reinforcement learning efficiency in shifted dynamics, suggesting a novel exploration strategy and better generalization.

CompFlow improves generalization for learned target dynamics, enhancing reinforcement learning processes.
The method constructs a conditional flow based on source-domain data and utilizes the wasserstein distance for estimating dynamics gaps.
Through optimal transport principles, CompFlow introduces a data collection strategy that prioritizes exploration in high dynamics gap areas.
The approach empirically demonstrates better performance across various reinforcement learning benchmarks, reducing the policy disparity.

Kong et al. (2025) studied this question.

View Full Paper

Ask AI

Bookmark

November 9, 2025Open Access

View Full Paper

Ask AI

Bookmark

Authors

CompFlow enhances reinforcement learning efficiency in shifted dynamics, suggesting a novel exploration strategy and better generalization.

CompFlow improves generalization for learned target dynamics, enhancing reinforcement learning processes.
The method constructs a conditional flow based on source-domain data and utilizes the wasserstein distance for estimating dynamics gaps.
Through optimal transport principles, CompFlow introduces a data collection strategy that prioritizes exploration in high dynamics gap areas.
The approach empirically demonstrates better performance across various reinforcement learning benchmarks, reducing the policy disparity.

Kong et al. (2025) studied this question.

Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context:

View Full Paper

Ask AI

Bookmark