TRAIL-IRL proposes crossed reward-temperature mixtures for unlabeled trajectory clustering
A new arXiv paper, TRAIL-IRL, introduces crossed reward-temperature mixtures for inverse reinforcement learning, noting that trajectories can differ in action randomness as well as underlying preferences.
Read the full story at arXiv stat.ML ↗
Timeline · 1 report
- 2026-11-18 05:00 · arXiv stat.ML
TRAIL-IRL: Crossed Reward-Temperature Mixtures for Unlabeled Trajectory Clustering