Physics-IQ benchmark reveals that generative video models exhibit limited physical understanding unrelated to their visual quality.
hub Mixed citations
How Far is Video Generation from World Model: A Physical Law Perspective
Mixed citation behavior. Most common role is background (62%).
40
Pith papers citing it
Background
62% of classified citations
abstract
OpenAI's Sora highlights the potential of video generation for developing world models that adhere to fundamental physical laws. However, the ability of video generation models to discover such laws purely from visual data without human priors can be questioned. A world model learning the true law should give predictions robust to nuances and correctly extrapolate on unseen scenarios. In this work, we evaluate across three key scenarios: in-distribution, out-of-distribution, and combinatorial generalization. We developed a 2D simulation testbed for object movement and collisions to generate videos deterministically governed by one or more classical mechanics laws. This provides an unlimited supply of data for large-scale experimentation and enables quantitative evaluation of whether the generated videos adhere to physical laws. We trained diffusion-based video generation models to predict object movements based on initial frames. Our scaling experiments show perfect generalization within the distribution, measurable scaling behavior for combinatorial generalization, but failure in out-of-distribution scenarios. Further experiments reveal two key insights about the generalization mechanisms of these models: (1) the models fail to abstract general physical rules and instead exhibit "case-based" generalization behavior, i.e., mimicking the closest training example; (2) when generalizing to new cases, models are observed to prioritize different factors when referencing training data: color > size > velocity > shape. Our study suggests that scaling alone is insufficient for video generation models to uncover fundamental physical laws, despite its role in Sora's broader success. See our project page at https://phyworld.github.io
hub tools
citation-role summary
background 6
baseline 2
citation-polarity summary
representative citing papers
None of ten tested video-generation models reliably remembers objects after occlusion in dynamic scenes; static-camera videos inflate consistency scores.
ImageTime is a benchmark that probes image generation models' visual world modeling by requiring coherent four-state sequences in single images, scored via VLM judge.
PhaseLock extracts motion priors from 2-step inference and enforces them via Latent Delta Guidance to raise physical consistency scores by 6.2 points on average in image-to-video diffusion models.
Dream.exe evaluates 8 video generation models on 101 manipulation tasks by converting generated videos into executable robot trajectories in a simulator, finding measurable success rates that visual metrics do not predict.
YoCausal benchmark shows video diffusion models detect the arrow of time but lack genuine causal understanding relative to humans.
CRONOS benchmark shows recent open-source video generators fail to preserve physical consistency under controlled changes to viewpoint, scene, object category, and appearance.
Aero-World adapts a pretrained latent diffusion transformer for action-conditioned aerial video generation by injecting inertial action tokens and using a frozen latent-space Physics Probe for inertial consistency supervision during LoRA finetuning, with a new AeroBench benchmark showing improved AA
WorldVLN proposes the first autoregressive world action model for aerial vision-language navigation that predicts short-horizon latent world states, decodes them to waypoints in closed loop, and uses two-stage training with Action-aware GRPO to achieve over 12% success-rate gains on benchmarks plus零
TeDiO regularizes temporal diagonals in diffusion transformer attention maps to produce smoother video motion while keeping per-frame quality intact.
ACWM-Phys is a controllable simulator benchmark with in- and out-of-distribution protocols for evaluating action-conditioned world models across rigid, kinematic, deformable, and particle dynamics.
NOVA represents world states as INR weights for decoder-free rendering, compactness, and unsupervised disentanglement of background, foreground, and motion in video world models.
NEO is a probabilistic neural model that induces compositional programs as a learned Language of Thought from non-textual observations and executes them via a shared transition model to enable explanation-driven generalization.
Do-Undo Bench is a new evaluation task and dataset that forces models to simulate forward action effects and then undo them to measure genuine action understanding in image generation.
DreamGen trains robot policies on synthetic trajectories from adapted video world models, enabling a humanoid robot to perform 22 new behaviors in seen and unseen environments from a single pick-and-place teleoperation dataset.
CrashTwin recovers metric-scale crash dynamics from monocular rollouts and shows that strong visual scores routinely mask large momentum, energy, and identity violations in world models.
Next Forcing augments video generation models with auxiliary multi-chunk prediction modules to achieve faster training convergence, higher accuracy at high frame rates, and 2x faster inference on world modeling benchmarks.
TunerDiT adds event-partitioned masking and cross-event prompt fusion to diffusion transformers for training-free multi-event video generation, with gains scaling by event count on a new Meve benchmark.
Generative sequence models for physical tasks exhibit physical misgeneralization where local prediction errors propagate through physical measurements to distort aggregate distributions over quantities like distance or energy; a data deviation kernel explains and predicts the shifts and supports a内核
NEWTON improves physical accuracy in video generation by deploying a trainable planner that coordinates physics-aware tools and a verifier, raising joint accuracy on VideoPhy-2 without altering the base generators.
MIGA introduces two-stage alignment to close train-inference gaps and dual consistency enhancement via self-reflection and long-range guidance to achieve SOTA temporal consistency in infinite-frame video generation on VBench and NarrLV.
VideoRLVR uses SDE-GRPO optimization, dense decomposed rewards, and Early-Step Focus to train video diffusion models on verifiable reasoning tasks, outperforming supervised fine-tuning and other video generators on Maze, FlowFree, and Sokoban.
The paper presents WorldReasonBench, a benchmark that tests video generators on maintaining physical, social, logical, and informational consistency when predicting future states from initial conditions and actions.
TRAP is a tail-aware ranking attack that plants a backdoor in world models so that a trigger causes the model to reorder a few critical imagined trajectories and redirect planning while preserving normal behavior on clean inputs.
citing papers explorer
-
Do generative video models understand physical principles?
Physics-IQ benchmark reveals that generative video models exhibit limited physical understanding unrelated to their visual quality.
-
MemoBench: Benchmarking World Modeling in Dynamically Changing Environments
None of ten tested video-generation models reliably remembers objects after occlusion in dynamic scenes; static-camera videos inflate consistency scores.
-
Can Image Models Imagine Time? ImageTime: A Novel Benchmark for Probing Visual World Modeling Through Spatiotemporal Consistency
ImageTime is a benchmark that probes image generation models' visual world modeling by requiring coherent four-state sequences in single images, scored via VLM judge.
-
Physics in 2-Steps: Locking Motion Priors Before Visual Refinement Erases Them
PhaseLock extracts motion priors from 2-step inference and enforces them via Latent Delta Guidance to raise physical consistency scores by 6.2 points on average in image-to-video diffusion models.
-
Dream.exe: Can Video Generation Models Dream Executable Robot Manipulation?
Dream.exe evaluates 8 video generation models on 101 manipulation tasks by converting generated videos into executable robot trajectories in a simulator, finding measurable success rates that visual metrics do not predict.
-
YoCausal: How Far is Video Generation from World Model? A Causality Perspective
YoCausal benchmark shows video diffusion models detect the arrow of time but lack genuine causal understanding relative to humans.
-
CRONOS: Benchmarking Counterfactual Physical Consistency in Video Models
CRONOS benchmark shows recent open-source video generators fail to preserve physical consistency under controlled changes to viewpoint, scene, object category, and appearance.
-
Aero-World: Action-Conditioned Aerial Video Generation from Inertial Controls
Aero-World adapts a pretrained latent diffusion transformer for action-conditioned aerial video generation by injecting inertial action tokens and using a frozen latent-space Physics Probe for inertial consistency supervision during LoRA finetuning, with a new AeroBench benchmark showing improved AA
-
WorldVLN: Autoregressive World Action Model for Aerial Vision-Language Navigation
WorldVLN proposes the first autoregressive world action model for aerial vision-language navigation that predicts short-horizon latent world states, decodes them to waypoints in closed loop, and uses two-stage training with Action-aware GRPO to achieve over 12% success-rate gains on benchmarks plus零
-
TeDiO: Temporal Diagonal Optimization for Training-Free Coherent Video Diffusion
TeDiO regularizes temporal diagonals in diffusion transformer attention maps to produce smoother video motion while keeping per-frame quality intact.
-
ACWM-Phys: Investigating Generalized Physical Interaction in Action-Conditioned Video World Models
ACWM-Phys is a controllable simulator benchmark with in- and out-of-distribution protocols for evaluating action-conditioned world models across rigid, kinematic, deformable, and particle dynamics.
-
Render, Don't Decode: Weight-Space World Models with Latent Structural Disentanglement
NOVA represents world states as INR weights for decoder-free rendering, compactness, and unsupervised disentanglement of background, foreground, and motion in video world models.
-
Learning to Theorize the World from Observation
NEO is a probabilistic neural model that induces compositional programs as a learned Language of Thought from non-textual observations and executes them via a shared transition model to enable explanation-driven generalization.
-
Do-Undo Bench: Reversibility for Action Understanding in Image Generation
Do-Undo Bench is a new evaluation task and dataset that forces models to simulate forward action effects and then undo them to measure genuine action understanding in image generation.
-
DreamGen: Unlocking Generalization in Robot Learning through Video World Models
DreamGen trains robot policies on synthetic trajectories from adapted video world models, enabling a humanoid robot to perform 22 new behaviors in seen and unseen environments from a single pick-and-place teleoperation dataset.
-
A Physics-Grounded Benchmark for Multi-Agent Dynamics in World Models
CrashTwin recovers metric-scale crash dynamics from monocular rollouts and shows that strong visual scores routinely mask large momentum, energy, and identity violations in world models.
-
Next Forcing: Causal World Modeling with Multi-Chunk Prediction
Next Forcing augments video generation models with auxiliary multi-chunk prediction modules to achieve faster training convergence, higher accuracy at high frame rates, and 2x faster inference on world modeling benchmarks.
-
TunerDiT: Training-free Progressive Steering of Diffusion Transformer for Multi-Event Video Generation
TunerDiT adds event-partitioned masking and cross-event prompt fusion to diffusion transformers for training-free multi-event video generation, with gains scaling by event count on a new Meve benchmark.
-
Mechanisms of Misgeneralization in Physical Sequence Modeling
Generative sequence models for physical tasks exhibit physical misgeneralization where local prediction errors propagate through physical measurements to distort aggregate distributions over quantities like distance or energy; a data deviation kernel explains and predicts the shifts and supports a内核
-
NEWTON: Agentic Planning for Physically Grounded Video Generation
NEWTON improves physical accuracy in video generation by deploying a trainable planner that coordinates physics-aware tools and a verifier, raising joint accuracy on VideoPhy-2 without altering the base generators.
-
Enhancing Train-Free Infinite-Frame Generation for Consistent Long Videos
MIGA introduces two-stage alignment to close train-inference gaps and dual consistency enhancement via self-reflection and long-range guidance to achieve SOTA temporal consistency in infinite-frame video generation on VBench and NarrLV.
-
Video Models Can Reason with Verifiable Rewards
VideoRLVR uses SDE-GRPO optimization, dense decomposed rewards, and Early-Step Focus to train video diffusion models on verifiable reasoning tasks, outperforming supervised fine-tuning and other video generators on Maze, FlowFree, and Sokoban.
-
WorldReasonBench: Human-Aligned Stress Testing of Video Generators as Future World-State Predictors
The paper presents WorldReasonBench, a benchmark that tests video generators on maintaining physical, social, logical, and informational consistency when predicting future states from initial conditions and actions.
-
TRAP: Tail-aware Ranking Attack for World-Model Planning
TRAP is a tail-aware ranking attack that plants a backdoor in world models so that a trigger causes the model to reorder a few critical imagined trajectories and redirect planning while preserving normal behavior on clean inputs.
-
When Do Diffusion Models learn to Generate Multiple Objects?
Using the mosaic controlled dataset framework, experiments show scene complexity dominates over concept imbalance in diffusion model failures for multi-object generation, with counting especially hard in low-data regimes and compositional generalization collapsing under held-out combinations.
-
Alice v1: Distillation-Enhanced Video Generation Surpassing Closed-Source Models
Alice v1 is an open video model that surpasses its teacher and closed-source systems like Veo3 and Sora2 in quality while running 7x faster through specialized distillation.
-
SceneScribe-1M: A Large-Scale Video Dataset with Comprehensive Geometric and Semantic Annotations
SceneScribe-1M is a new dataset of 1 million videos with semantic text, camera parameters, dense depth, and consistent 3D point tracks to support monocular depth estimation, scene reconstruction, point tracking, and text-to-video synthesis.
-
VisPhyWorld: Probing Physical Reasoning via Code-Driven Video Reconstruction
VisPhyWorld evaluates MLLMs' physical reasoning via executable code generation for video reconstruction, with VisPhyBench showing strong semantics but weak parameter inference and dynamics simulation.
-
ProPhy: Progressive Physical Alignment for Dynamic World Simulation
ProPhy adds explicit physics-aware conditioning via semantic and refinement experts plus VLM knowledge transfer to produce more physically coherent dynamic videos than prior methods.
-
Video models are zero-shot learners and reasoners
Generative video models exhibit emergent zero-shot capabilities across perception, manipulation, and basic reasoning tasks.
-
ReSim: Reliable World Simulation for Autonomous Driving
ReSim is a controllable video world model trained on heterogeneous real and simulated driving data that achieves higher fidelity and controllability for both expert and non-expert actions, plus a Video2Reward module for estimating action quality from simulated futures.
-
A Definition and Roadmap for World Models
A perspective article defining world models as finite-resource compression of physical state transitions and outlining a roadmap toward physical AGI via unified representations and interactive simulators.
-
PhysRAG: Enhancing Physics-Awareness in Video Generation via Retrieval-Augmented Generation
PhysRAG curates 7K videos from WISA-80K, builds a physical video database, and injects knowledge via learnable queries into a diffusion model to reach SOTA visual quality and physical compliance on PhyGenBench and VBench.
-
Physics-IQ Verified
Physics-IQ Verified refines 57.6% of samples and 34.8% of prompts from the original benchmark and produces moderate ranking shifts (Kendall's τ = 0.46) across six image-to-video models.
-
MPMWorlds: Material-Point-Method Simulations for Inferring and Extrapolating Physical Dynamics
Assembles MPM simulation dataset and compares code generation versus video diffusion for inferring physical parameters and extrapolating dynamics from videos.
-
Physically Viable World Models: A Case for Query-Conditioned Embodied AI
Embodied AI requires query-conditioned world models that select the simplest physical abstraction sufficient to answer intervention queries.
-
MASS: Motion-Aware Spatial-Temporal Grounding for Physics Reasoning and Comprehension in Vision-Language Models
MASS adds spatiotemporal motion signals and 3D grounding to VLMs and releases MASS-Bench, yielding physics-reasoning performance within 2% of Gemini-2.5-Flash after reinforcement fine-tuning.
-
Cloning Deterministic Worlds: The Critical Role of Latent Geometry in Long-Horizon World Models
GRWM uses temporal contrastive learning to geometrically regularize latent spaces in world models for high-fidelity cloning of deterministic 3D worlds.
-
Cosmos World Foundation Model Platform for Physical AI
The Cosmos platform supplies open-source pre-trained world models and supporting tools for building fine-tunable digital world simulations to train Physical AI.
-
NVIDIA Isaac Sim: Enabling Scalable, GPU-Accelerated Simulation for Robotics
A survey reviewing the architecture, usage patterns, and limitations of NVIDIA Isaac Sim across robotics domains.