mlsanityData to 14 Oct 2026

Discovering What You Can Control: Interventional Boundary Discovery for Reinforcement Learning

arXiv 2603.18257 ↗First posted 18 Mar 2026cs.LG cs.AIcited by 2

On the map: Deep reinforcement learning methods → Deep RL algorithms and representations

Abstract (arXiv)

When an RL agent's observations contain distractors driven by the same confounders as its true state, observational data alone cannot identify which dimensions the agent controls. In our benchmarks, even state-conditioned observational selectors can collapse when distractors mimic controllable state variables. We propose Interventional Boundary Discovery (IBD), which treats the agent's own action channel as a source of randomized interventions: randomizing actions implements an interventional contrast, and per-dimension two-sample tests with FDR correction produce a binary mask over observation dimensions. Across 12 continuous-control settings with up to 100 distractors, IBD matches oracle return in 11 of 12 settings, while observational baselines including mutual information, state-conditioned forward models, and gradient-based sensitivity often underperform simply passing the full observation to SAC. Code is available at https://github.com/jiaxin26/IBD-RL

Key facts. Cited by 2 papers in the map (data through 14 Oct 2026). Placed in Deep RL algorithms and representations, one of 15 directions in Deep reinforcement learning methods. Builds on 23 of its 34 references that are in the map. First posted on arXiv 18 Mar 2026.

Builds on

23 of 34 references.

Closest ancestors · 3

Same direction, newest first

  1. 2025Confounding Robust Deep Reinforcement Learning: A Causal Approach24 Oct
  2. 2025A Forget-and-Grow Strategy for Deep Reinforcement Learning Scaling in Continuous Control3 Jul
  3. 2025Zero-Shot Reinforcement Learning Under Partial Observability18 Jun

Other references · 20

  1. 2018Soft Actor-Critic: Off-Policy Maximum Entropy Deep Reinforcement Learning with a Stochastic Actorcited by 4,885
  2. 2018Addressing Function Approximation Error in Actor-Critic Methodscited by 2,417
  3. 2018DeepMind Control Suitecited by 956
  4. 2020CURL: Contrastive Unsupervised Representations for Reinforcement Learningcited by 760
  5. 2021Mastering Visual Continuous Control: Improved Data-Augmented Reinforcement Learningcited by 372
  6. 2015Variational Information Maximisation for Intrinsically Motivated Reinforcement Learningcited by 144
  7. 2022CITRIS: Causal Identifiability from Temporal Intervened Sequencescited by 133
  8. 2021Stabilizing Deep Q-Learning with ConvNets and Vision Transformers under Data Augmentationcited by 105
  9. 2021The Distracting Control Suite -- A Challenging Benchmark for Reinforcement Learning from Pixelscited by 100
  10. 2022Denoised MDPs: Learning World Models Better Than the World Itselfcited by 68
  11. 2024WoCoCo: Learning Whole-Body Humanoid Control with Sequential Contactscited by 61
  12. 2022Causal Representation Learning for Instantaneous and Temporal Effects in Interactive Systemscited by 51
8 more
  1. 2024Unifying Causal Representation Learning with the Invariance Principlecited by 31
  2. 2021Provable RL with Exogenous Distractors via Multistep Inverse Dynamicscited by 28
  3. 2024Score-based Causal Representation Learning: Linear and General Transformationscited by 27
  4. 2024Linear Causal Representation Learning from Unknown Multi-node Interventionscited by 14
  5. 2025PIGDreamer: Privileged Information Guided World Models for Safe Partially Observable Reinforcement Learningcited by 7
  6. 2024Learning a Fast Mixing Exogenous Block MDP using a Single Trajectorycited by 3
  7. 2022Disentangled (Un)Controllable Featurescited by 2
  8. 2025Offline Action-Free Learning of Ex-BMDPs by Comparing Diverse Datasetscited by 2

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2026: 2.

2 papers.

  1. 2026RAM: Recover Any 3D Human Motion in-the-Wildcited by 12
  2. 2026BCL: Bayesian In-Context Learning Framework for Information Extractioncited by 1

Similar, not cited

Close in wording, no citation either way

  1. 2026Why LLMs Fail at Causal Discovery and How Interventional Agents Escape
  2. 2024Graph-attention-based Casual Discovery with Trust Region-navigated Clipping Policy Optimization
  3. 2026Causal Retention in Interactive Agents: Interface Factorization and Selective Adaptation
  4. 2026Agentic Reinforcement Learning with Observation-Calibrated Self-Distillation
  5. 2025From Pixels to Factors: Learning Independently Controllable State Variables for Reinforcement Learning
  6. 2025DISCOVER: Automated Curricula for Sparse-Reward Reinforcement Learning
  7. 2023Initial State Interventions for Deconfounded Imitation Learning
  8. 2026AlphaG-OPD: Reliability-Gated Sibling Counterfactuals for On-Policy Distillation in Symbolic Alpha Factor Discovery