aiqbench analysis
This research introduces ConTraIRL, a novel framework in inverse reinforcement learning that decouples environment dynamics and task goals into latent representations to enable robust reward transfer across unseen scenarios. By utilizing a dual-encoder architecture with contrastive objectives, the method significantly improves sample efficiency and generalization on continuous control benchmarks compared to existing baselines. The work advances the state of machine learning by addressing critical reliability issues in policy transfer when facing novel combinations of dynamics and objectives.
ConTraIRL: Factorized Contrastive Abstractions for Transferable IRL
arXiv:2606.03017v2 Announce Type: replace Abstract: Reward transfer in Inverse Reinforcement Learning (IRL) is unreliable when policies must generalize to unseen combinations of environment dynamics and task goals. We propose Factorized Contrastive Abstractions for Transferable IRL (ConTraIRL), a framework that enables compositional reward transfer by learning decoupled latent representations of these two factors. ConTraIRL uses a dual-encoder architecture that maps observations into separate dynamics and goal latent spaces, trained with a dual contrastive objective. Temporal alignment encourages the dynamics encoder to learn goal-invariant structure, while the goal encoder captures dynamics-invariant features. This factorization supports reward inference under recombined dynamics-goal settings. Experiments on continuous control benchmarks demonstrate effective few-shot transfer to unseen dynamics-goal pairings, improving sample efficiency and reward recovery over transfer IRL baselines.
Abstract:Reward transfer in Inverse Reinforcement Learning (IRL) is unreliable when policies must generalize to unseen combinations of environment dynamics and task goals. We propose Factorized Contrastive Abstractions for Transferable IRL (ConTraIRL), a framework that enables compositional reward transfer by learning decoupled latent representations of these two factors. ConTraIRL uses a dual-encoder architecture that maps observations into separate dynamics and goal latent spaces, trained with a dual contrastive objective. Temporal alignment encourages the dynamics encoder to learn goal-invariant structure, while the goal encoder captures dynamics-invariant features. This factorization supports reward inference under recombined dynamics-goal settings. Experiments on continuous control benchmarks demonstrate effective few-shot transfer to unseen dynamics-goal pairings, improving sample efficiency and reward recovery over transfer IRL baselines.
Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Robotics (cs.RO)
arXiv:2606.03017 [cs.LG]
(or arXiv:2606.03017v2 [cs.LG] for this version)
https://doi.org/10.48550/arXiv.2606.03017 Focus to learn more arXiv-issued DOI via DataCite
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Full text under the publisher's open licence · 214 words
Related coverage
ConTraIRL: Factorized Contrastive Abstractions for Transferable IRL
arXiv · Jun 3, 2026
Game Design Creativity evals
LessWrong · Oct 12, 2026
Learning Cross-Model Activation Alignments with Explicit Many-to-Many Layer Maps
arXiv · Oct 8, 2026
Probabilistic Contrastive Pretraining for Multi-task ADME Property Prediction
arXiv · Jun 12, 2026