mlsanityData to 7 Oct 2026

Rapid trial-and-error learning with simulation supports flexible tool use and physical reasoning

arXiv 1907.09620 ↗First posted 22 Jul 2019cs.AI cs.LG cs.ROcited by 60

Not on the map. Why?

Abstract (arXiv)

Many animals, and an increasing number of artificial agents, display sophisticated capabilities to perceive and manipulate objects. But human beings remain distinctive in their capacity for flexible, creative tool use -- using objects in new ways to act on the world, achieve a goal, or solve a problem. To study this type of general physical problem solving, we introduce the Virtual Tools game. In this game, people solve a large range of challenging physical puzzles in just a handful of attempts. We propose that the flexibility of human physical problem solving rests on an ability to imagine the effects of hypothesized actions, while the efficiency of human search arises from rich action priors which are updated via observations of the world. We instantiate these components in the "Sample, Simulate, Update" (SSUP) model and show that it captures human performance across 30 levels of the Virtual Tools game. More broadly, this model provides a mechanism for explaining how people condense general physical knowledge into actionable, task-specific plans to achieve flexible and efficient physical problem-solving.

Key facts. Cited by 60 papers in the map (data through 7 Oct 2026). First posted on arXiv 22 Jul 2019.

Builds on

No reference list yet.

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2021: 9; 2022: 13; 2023: 8; 2024: 11; 2025: 11; 2026: 8.

60 papers.

Sort:

By area:

  1. 2023Tool Learning with Foundation Modelscited by 302
  2. 2024How Far is Video Generation from World Model: A Physical Law Perspectivecited by 204
  3. 2021Multi-Task Reinforcement Learning with Context-based Representationscited by 169
  4. 2021Physion: Evaluating Physical Prediction from Vision in Humans and Machinescited by 145
  5. 2024Pandora: Towards General World Model with Natural Language Actions and Video Statescited by 83
  6. 2024Building Machines that Learn and Think with Peoplecited by 73
  7. 2021Unsupervised Learning of Compositional Energy Conceptscited by 71
  8. 2024Towards Lifelong Learning of Large Language Models: A Surveycited by 60
  9. 2023Language Models, Agent Models, and World Models: The LAW for Machine Reasoning and Planningcited by 59
  10. 2021Human-Level Reinforcement Learning through Theory-Based Modeling, Exploration, and Planningcited by 43
  11. 2022Learning rigid dynamics with face interaction graph networkscited by 43
  12. 2024PoCo: Policy Composition from and for Heterogeneous Robot Learningcited by 42
38 more
  1. 2024m&m's: A Benchmark to Evaluate Tool-Use for multi-step multi-modal Taskscited by 38
  2. 2024Core Knowledge Deficits in Multi-Modal Language Modelscited by 38
  3. 2022Physical Design using Differentiable Learned Simulatorscited by 37
  4. 2022A Review of Emerging Research Directions in Abstract Visual Reasoningcited by 33
  5. 2021Capturing the objects of vision with neural networkscited by 21
  6. 2025Assessing Adaptive World Models in Machines with Novel Gamescited by 20
  7. 2025TRAVL: A Recipe for Making Video-Language Models Better Judges of Physics Implausibilitycited by 16
  8. 2023I-PHYRE: Interactive Physical Reasoningcited by 15
  9. 2022Towards Understanding How Machines Can Learn Causal Overhypothesescited by 13
  10. 2024Compositional Image Decomposition with Diffusion Modelscited by 12
  11. 2022On the Learning Mechanisms in Physical Reasoningcited by 11
  12. 2021Phy-Q as a measure for physical reasoning intelligencecited by 10
  13. 2022A Thermodynamics-informed Active Learning Approach to Perception and Reasoning about Fluidscited by 9
  14. 2025VLMgineer: Vision Language Models as Robotic Toolsmithscited by 9
  15. 2023Benchmarks for Physical Reasoning AIcited by 8
  16. 2022Learning Causal Overhypotheses through Exploration in Children and Computational Modelscited by 7
  17. 2022Understanding Physical Effects for Effective Tool-usecited by 7
  18. 2023Learning to Design and Use Tools for Robotic Manipulationcited by 7
  19. 2023diff History for Neural Language Agentscited by 6
  20. 2024PhyPlan: Compositional and Adaptive Physical Task Reasoning with Physics-Informed Skill Networks for Robot Manipulatorscited by 6
  21. 2026Learning Physics-Grounded 4D Dynamics with Neural Gaussian Force Fieldscited by 5
  22. 2025RobotSmith: Generative Robotic Tool Design for Acquisition of Complex Manipulation Skillscited by 4
  23. 2022H-SAUR: Hypothesize, Simulate, Act, Update, and Repeat for Understanding Object Articulations from Interactionscited by 3
  24. 2022Discovering Generalizable Spatial Goal Representations via Graph-based Active Reward Learningcited by 3
  25. 2025Afford-X: Generalizable and Slim Affordance Reasoning for Task-oriented Manipulationcited by 3
  26. 2026MCPShield: A Security Cognition Layer for Adaptive Trust Calibration in Model Context Protocol Agentscited by 3
  27. 2021Visual scoping operations for physical assemblycited by 2
  28. 2023NovPhy: A Testbed for Physical Reasoning in Open-world Environmentscited by 2
  29. 2025Revisiting Rogers' Paradox in the Context of Human-AI Interactioncited by 2
  30. 2025Causal-PIK: Causality-based Physical Reasoning with a Physics-Informed Kernelcited by 2
  31. 2026Hierarchical Experimentalist Agentscited by 2
  32. 2021The Difficulty of Novelty Detection in Open-World Physical Domains: An Application to Angry Birdscited by 1
  33. 2021Hi-Phy: A Benchmark for Hierarchical Physical Reasoningcited by 1
  34. 2022Modeling Human Behavior Part I -- Learning and Belief Approachescited by 1
  35. 2024Grounding Large Language Models In Embodied Environment With Imperfect World Modelscited by 1
  36. 2025Neural Force Field: Few-shot Learning of Generalized Physical Reasoningcited by 1
  37. 2025A simulation-heuristics dual-process model for intuitive physicscited by 1
  38. 2025Mini Amusement Parks (MAPs): A Testbed for Modelling Business Decisionscited by 1

50 most cited of 60 All 60 (JSON)