mlsanityData to 7 Oct 2026

CausalVAE as a Plug-in for World Models: Towards Reliable Counterfactual Dynamics

arXiv 2604.07712 ↗First posted 9 Apr 2026cs.LGcited by 2

On the map: World models → World models for language agents

Abstract (arXiv)

In this work, CausalVAE is introduced as a plug-in structural module for latent world models and is attached to diverse encoder-transition backbones. Across the reported benchmarks, competitive factual prediction is preserved and intervention-aware counterfactual retrieval is improved after the plug-in is added, suggesting stronger robustness under distribution shift and interventions. The largest gains are observed on the Physics benchmark: when averaged over 8 paired baselines, CF-H@1 is improved by +102.5%. In a representative GNN-NLL setting on Physics, CF-H@1 is increased from 11.0 to 41.0 (+272.7%). Through causal analysis, learned structural dependencies are shown to recover meaningful first-order physical interaction trends, supporting the interpretability of the learned latent causal structure.

Key facts. Cited by 2 papers in the map (data through 7 Oct 2026). Placed in World models for language agents, one of 11 directions in World models. Builds on 16 of its 21 references that are in the map. First posted on arXiv 9 Apr 2026.

Builds on

16 of 21 references.

Closest ancestors · 1

Same direction, newest first

  1. 2020Causal World Models by Unsupervised Deconfounding of Physical Dynamics28 Dec

Other references · 15

  1. 2019Dream to Control: Learning Behaviors by Latent Imaginationcited by 1,598
  2. 2018World Modelscited by 1,444
  3. 2023Mastering Diverse Domains through World Modelscited by 1,193
  4. 2020Mastering Atari with Discrete World Modelscited by 990
  5. 2023Self-Supervised Learning from Images with a Joint-Embedding Predictive Architecturecited by 982
  6. 2020Learning to Simulate Complex Physics with Graph Networkscited by 867
  7. 2020Object-Centric Learning with Slot Attentioncited by 865
  8. 2018DAGs with NO TEARS: Continuous Optimization for Structure Learningcited by 745
  9. 2019Variational Autoencoders and Nonlinear ICA: A Unifying Frameworkcited by 494
  10. 2024Revisiting Feature Prediction for Learning Visual Representations from Videocited by 490
  11. 2019MONet: Unsupervised Scene Decomposition and Representationcited by 388
  12. 2019Multi-Object Representation Learning with Iterative Variational Inferencecited by 359
3 more
  1. 2018Neural Relational Inference for Interacting Systemscited by 339
  2. 2021Systematic Evaluation of Causal Discovery in Visual Model Based Reinforcement Learningcited by 53
  3. 2023Understanding Physical Dynamics with Counterfactual World Modelingcited by 11

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2026: 2.

2 papers.

  1. 2026Latent State Design for World Models under Sufficiency Constraintscited by 3
  2. 2026Event-Conditioned Diagnostics of Kinematic, Contact, and Object-Permanence Structure in Passive Object-State World Modelscited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2025The Causal Round Trip: Generating Authentic Counterfactuals by Eliminating Information Loss
  2. 2022Variational Causal Dynamics: Discovering Modular World Models from Interventions
  3. 2025Inducing Causal World Models in LLMs for Zero-Shot Physical Reasoning
  4. 2023Modeling Causal Mechanisms with Diffusion Models for Interventional and Counterfactual Queries
  5. 2024CausalBN-Bench: A Comprehensive Benchmark for Causal Learning Capability of LLMs
  6. 2026Causal Scaffolding for Physical Reasoning: A Benchmark for Causally-Informed Physical World Understanding in VLMs
  7. 2026CausalWM: Causal Chain-of-Thought Reasoning for Embodied World Model
  8. 2026Low-Rank Dynamics-Effective Latent Carriers for Counterfactual Rollout in Learned World Models