mlsanityData to 7 Oct 2026

Intuitive physics understanding emerges from self-supervised pretraining on natural videos

arXiv 2502.11831 ↗First posted 17 Feb 2025cs.CV cs.AIcited by 66

Not on the map. Why?

Cited by 25 in World models, but placed by its own abstract, which didn’t fit.

Abstract (arXiv)

We investigate the emergence of intuitive physics understanding in general-purpose deep neural network models trained to predict masked regions in natural videos. Leveraging the violation-of-expectation framework, we find that video prediction models trained to predict outcomes in a learned representation space demonstrate an understanding of various intuitive physics properties, such as object permanence and shape consistency. In contrast, video prediction in pixel space and multimodal large language models, which reason through text, achieve performance closer to chance. Our comparisons of these architectures reveal that jointly learning an abstract representation space while predicting missing parts of sensory input, akin to predictive coding, is sufficient to acquire an understanding of intuitive physics, and that even models trained on one week of unique video achieve above chance performance. This challenges the idea that core knowledge -- a set of innate systems to help understand the world -- needs to be hardwired to develop an understanding of intuitive physics.

Key facts. Cited by 66 papers in the map (data through 7 Oct 2026). First posted on arXiv 17 Feb 2025.

Builds on

0 of 0 references.

No references here yet.

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2025: 24; 2026: 42.

66 papers.

Sort:

By area:

  1. 2026LeWorldModel: Stable End-to-End Joint-Embedding Predictive Architecture from Pixelscited by 237
  2. 2025Video models are zero-shot learners and reasonerscited by 208
  3. 2025Cambrian-S: Towards Spatial Supersensing in Videocited by 123
  4. 2026V-JEPA 2.1: Unlocking Dense Features in Video Self-Supervised Learningcited by 104
  5. 2025VideoREPA: Learning Physics for Video Generation through Relational Alignment with Foundation Modelscited by 88
  6. 2025Back to the Features: DINO as a Foundation for Video World Modelscited by 56
  7. 2025IntPhys 2: Benchmarking Intuitive Physics Understanding In Complex Synthetic Environmentscited by 49
  8. 2025NewtonGen: Physics-Consistent and Controllable Text-to-Video Generation via Neural Newtonian Dynamicscited by 40
  9. 2026Inference-time Physics Alignment of Video Generative Models with Latent World Modelscited by 32
  10. 2025A Shortcut-aware Video-QA Benchmark for Physical Understanding via Minimal Video Pairscited by 30
  11. 2025What about gravity in video generation? Post-Training Newton's Laws with Verifiable Rewardscited by 23
  12. 2026JEPA-VLA: Video Predictive Embedding is Needed for VLA Modelscited by 22
38 more
  1. 2025WorldPrediction: A Benchmark for High-level World Modeling and Long-horizon Procedural Planningcited by 19
  2. 2026Interpreting Physics in Video World Modelscited by 18
  3. 2025Synthetic Video Enhances Physical Fidelity in Video Synthesiscited by 15
  4. 2025LikePhys: Evaluating Intuitive Physics Understanding in Video Diffusion Models via Likelihood Preferencecited by 14
  5. 2026A Lightweight Library for Energy-Based Joint-Embedding Predictive Architecturescited by 13
  6. 2025Rethinking JEPA: Compute-Efficient Video SSL with Frozen Teacherscited by 11
  7. 2026Phantom: Physics-Infused Video Generation via Joint Modeling of Visual and Latent Physical Dynamicscited by 10
  8. 2026NTIRE 2026 Challenge on Video Saliency Prediction: Methods and Resultscited by 10
  9. 2026Why AI systems don't learn and what to do about it: Lessons on autonomous learning from cognitive sciencecited by 8
  10. 2025seq-JEPA: Autoregressive Predictive Learning of Invariant-Equivariant World Modelscited by 7
  11. 2025Seeing the Wind from a Falling Leafcited by 7
  12. 2026DiLA: Disentangled Latent Action World Modelscited by 6
  13. 2025Diffusion-based 3D Hand Motion Recovery with Intuitive Physicscited by 5
  14. 2025From 2D to 3D Cognition: A Brief Survey of General World Modelscited by 4
  15. 2026From Generative Engines to Actionable Simulators: The Imperative of Physical Grounding in World Modelscited by 4
  16. 2026LaMo: Self-Supervised Latent Motion Priors for Physical Realism in Video Generationcited by 4
  17. 2026The Invisible Hand of Physics: When Video Diffusion Models Know More Than They Showcited by 4
  18. 2025World Models for Anomaly Detection during Model-Based Reinforcement Learning Inferencecited by 3
  19. 2025Zero-Shot Visual Generalization in Robot Manipulationcited by 3
  20. 2025Objects in Generated Videos Are Slower Than They Appear: Models Suffer Sub-Earth Gravity and Don't Know Galileo's Principle...for nowcited by 3
  21. 2025SeeU: Seeing the Unseen World via 4D Dynamics-aware Generationcited by 3
  22. 2026VisPhyWorld: Probing Physical Reasoning via Code-Driven Video Reconstructioncited by 3
  23. 2026How VLAs (Really) Work In Open-World Environmentscited by 3
  24. 2026Latent Video Prediction Learns Better World Modelscited by 3
  25. 2026YoCausal: How Far is Video Generation from World Model? A Causality Perspectivecited by 3
  26. 2026What Can Latent World Models Know? Physical Parameter Identifiability in Multimodal Predictive Representationscited by 3
  27. 2025Pixels to Principles: Probing Intuitive Physics Understanding in Multimodal Language Modelscited by 2
  28. 2025InPhyRe Discovers: Large Multimodal Models Struggle in Inductive Physical Reasoningcited by 2
  29. 2026GEOPHYS: The Geometry of Physical Plausibilitycited by 2
  30. 2026Self-Supervised Learning from Structural Invariancecited by 1
  31. 2026Seeking Physics in Diffusion Noisecited by 1
  32. 2026Emergent Compositional Communication for Latent World Propertiescited by 1
  33. 2026How Do Video Foundation Models Encode Intuitive Physics? Probing Across Pretraining Paradigmscited by 1
  34. 2026Apple-π: Benchmarking Thinking with Video Towards Law-Grounded Physical Intelligencecited by 1
  35. 2026PAVXploreRL: Physical-Action-Visual World Model Reinforcement Learning with Action Explorationcited by 1
  36. 2026Low-Rank Dynamics-Effective Latent Carriers for Counterfactual Rollout in Learned World Modelscited by 1
  37. 2025Towards aligned body representations in vision modelscited by 0
  38. 2025Opinion: Learning Intuitive Physics May Require More than Visual Datacited by 0

50 most cited of 66 All 66 (JSON)

Similar, not cited

Close in wording, no citation either way

  1. 2026Causal Physics Steering in Video World Models via Concept Activation Vectors
  2. 2023Visual cognition in multimodal large language models
  3. 2026Can Vision Language Models Learn Intuitive Physics from Interaction?
  4. 2025TRAVL: A Recipe for Making Video-Language Models Better Judges of Physics Implausibility
  5. 2026From Priors to Perception: Grounding Video-LLMs in Physical Reality
  6. 2026PhysMRV: Physical Memory Retrieval and Verification for Physics Plausibility Reasoning
  7. 2025Uncovering Spontaneous Physics Representations in In-Context Learning
  8. 2025Physics Knowledge in Frontier Models: A Diagnostic Study of Failure Modes