TROFI learns a reward model from ranked trajectories, labels an offline dataset with it, and trains a TD3+BC policy, matching ground-truth-reward performance on many D4RL tasks without a hand-coded reward or expert demonstrations.
It is an open-world city simulation originally proposed by Sestini et al
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
TROFI: Trajectory-Ranked Offline Inverse Reinforcement Learning
TROFI learns a reward model from ranked trajectories, labels an offline dataset with it, and trains a TD3+BC policy, matching ground-truth-reward performance on many D4RL tasks without a hand-coded reward or expert demonstrations.