Saved in:
| Main Authors: | , , |
|---|---|
| Format: | Preprint |
| Published: |
2026
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2601.19452 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866917225158410240 |
|---|---|
| author | Rietz, Finn Martires, Pedro Zuidberg dos Stork, Johannes Andreas |
| author_facet | Rietz, Finn Martires, Pedro Zuidberg dos Stork, Johannes Andreas |
| contents | Incorporating demonstration data into reinforcement learning (RL) can greatly accelerate learning, but existing approaches often assume demonstrations are optimal and fully aligned with the target task. In practice, demonstrations are frequently sparse, suboptimal, or misaligned, which can degrade performance when these demonstrations are integrated into RL. We propose Adaptive Policy Composition (APC), a hierarchical model that adaptively composes multiple data-driven Normalizing Flow (NF) priors. Instead of enforcing strict adherence to the priors, APC estimates each prior's applicability to the target task while leveraging them for exploration. Moreover, APC either refines useful priors, or sidesteps misaligned ones when necessary to optimize downstream reward. Across diverse benchmarks, APC accelerates learning when demonstrations are aligned, remains robust under severe misalignment, and leverages suboptimal demonstrations to bootstrap exploration while avoiding performance degradation caused by overly strict adherence to suboptimal demonstrations. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2601_19452 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | APC-RL: Exceeding Data-Driven Behavior Priors with Adaptive Policy Composition Rietz, Finn Martires, Pedro Zuidberg dos Stork, Johannes Andreas Machine Learning Artificial Intelligence Incorporating demonstration data into reinforcement learning (RL) can greatly accelerate learning, but existing approaches often assume demonstrations are optimal and fully aligned with the target task. In practice, demonstrations are frequently sparse, suboptimal, or misaligned, which can degrade performance when these demonstrations are integrated into RL. We propose Adaptive Policy Composition (APC), a hierarchical model that adaptively composes multiple data-driven Normalizing Flow (NF) priors. Instead of enforcing strict adherence to the priors, APC estimates each prior's applicability to the target task while leveraging them for exploration. Moreover, APC either refines useful priors, or sidesteps misaligned ones when necessary to optimize downstream reward. Across diverse benchmarks, APC accelerates learning when demonstrations are aligned, remains robust under severe misalignment, and leverages suboptimal demonstrations to bootstrap exploration while avoiding performance degradation caused by overly strict adherence to suboptimal demonstrations. |
| title | APC-RL: Exceeding Data-Driven Behavior Priors with Adaptive Policy Composition |
| topic | Machine Learning Artificial Intelligence |
| url | https://arxiv.org/abs/2601.19452 |