Dynamical Priors as a Training Objective in Reinforcement Learning
Fuente:
arXiv
Saved in:
| Main Author: | |
|---|---|
| Format: | Preprint |
| Published: |
2026
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866918464244940800 |
|---|---|
| author | Subaharan, Sukesh |
| author_facet | Subaharan, Sukesh |
| contents | Standard reinforcement learning (RL) optimizes policies for reward but imposes few constraints on how decisions evolve over time. As a result, policies may achieve high performance while exhibiting temporally incoherent behavior such as abrupt confidence shifts, oscillations, or degenerate inactivity. We introduce Dynamical Prior Reinforcement Learning (DP-RL), a training framework that augments policy gradient learning with an auxiliary loss derived from external state dynamics that implement evidence accumulation and hysteresis. Without modifying the reward, environment, or policy architecture, this prior shapes the temporal evolution of action probabilities during learning. Across three minimal environments, we show that dynamical priors systematically alter decision trajectories in task-dependent ways, promoting temporally structured behavior that cannot be explained by generic smoothing. These results demonstrate that training objectives alone can control the temporal geometry of decision-making in RL agents. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2604_21464 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Dynamical Priors as a Training Objective in Reinforcement Learning Subaharan, Sukesh Machine Learning Artificial Intelligence 68T05 I.2.6; I.2.11 Standard reinforcement learning (RL) optimizes policies for reward but imposes few constraints on how decisions evolve over time. As a result, policies may achieve high performance while exhibiting temporally incoherent behavior such as abrupt confidence shifts, oscillations, or degenerate inactivity. We introduce Dynamical Prior Reinforcement Learning (DP-RL), a training framework that augments policy gradient learning with an auxiliary loss derived from external state dynamics that implement evidence accumulation and hysteresis. Without modifying the reward, environment, or policy architecture, this prior shapes the temporal evolution of action probabilities during learning. Across three minimal environments, we show that dynamical priors systematically alter decision trajectories in task-dependent ways, promoting temporally structured behavior that cannot be explained by generic smoothing. These results demonstrate that training objectives alone can control the temporal geometry of decision-making in RL agents. |
| title | Dynamical Priors as a Training Objective in Reinforcement Learning |
| topic | Machine Learning Artificial Intelligence 68T05 I.2.6; I.2.11 |
| url | https://arxiv.org/abs/2604.21464 |