PAWBench: How Far Are We from Probabilistically Aligned World Modeling?

• URL: http://arxiv.org/abs/2608.27345v3
• Date: Thu, 03 Sep 2026 08:39:17 GMT
• Title: PAWBench: How Far Are We from Probabilistically Aligned World Modeling?
• Abstract summary: We call this distribution-level requirement probabilistic alignment.<n>PAWBench is a benchmark for evaluating video generators as samplers of world dynamics.<n>We introduce PAWEval, an outcome-level protocol that converts repeated video rollouts into empirical distributions.
• Score: 60.16773899025656
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Recent video generation models are increasingly framed as world models. Many physical processes can unfold in more than one valid way. Therefore, a world model should reproduce not only a plausible trajectory, but also the distribution of possible behaviors under the same initial observation and action. We call this distribution-level requirement probabilistic alignment. However, existing evaluations largely assess individual-video plausibility and do not test whether repeated generations recover the correct distribution. This raises a central question: how far are current video generators from probabilistically aligned world modeling? To answer it, we formalize probabilistic alignment as a distributional criterion for world models and introduce PAWBench, a benchmark for evaluating video generators as stochastic samplers of world dynamics. We further introduce PAWEval, an outcome-level protocol that converts repeated video rollouts into empirical distributions over possible physical behaviors. Across 50 scenarios and eleven current systems, no model consistently matches the reference probabilities while recovering the range of valid behaviors. Having established this gap, we test whether language prompts, initial noise sampling, or model training can reshape the model's predictive distribution. We believe our work can serve as a foundation for future efforts to move towards probabilistically aligned world modeling.

Related papers

• A Probabilistic Perspective on Model Collapse [9.087950471621653]
  This paper aims to characterize the conditions under which model collapse occurs and, crucially, how it can be mitigated.<n>Under mild conditions, we rigorously show that progressively increasing the sample size at each training step is necessary to prevent model collapse.<n>We also investigate the probability that training on synthetic data yields models that outperform those trained solely on real data.
  arXiv  Detail & Related papers  (2025-05-20T05:25:29Z)
• How Far is Video Generation from World Model: A Physical Law Perspective [101.24278831609249]
  OpenAI's Sora highlights the potential of video generation for developing world models that adhere to physical laws.<n>But the ability of video generation models to discover such laws purely from visual data without human priors can be questioned.<n>In this work, we evaluate across three key scenarios: in-distribution, out-of-distribution, and generalization.
  arXiv  Detail & Related papers  (2024-11-04T18:53:05Z)
• Learning Chaotic Systems and Long-Term Predictions with Neural Jump ODEs [4.204990010424083]
  Pathdependent Neural Jump ODE (PDNJ-ODE) is a model for online prediction of generic processes with irregular (in time) and potentially incomplete (with respect to coordinates) observations. In this work we enhance the model with two novel ideas, which independently of each other improve the performance of our modelling setup. The same enhancements can be used to provably enable the PDNJ-ODE to learn long-term predictions for general datasets, where the standard model fails.
  arXiv  Detail & Related papers  (2024-07-26T15:18:29Z)
• SimPro: A Simple Probabilistic Framework Towards Realistic Long-Tailed Semi-Supervised Learning [49.94607673097326]
  We propose a highly adaptable framework, designated as SimPro, which does not rely on any predefined assumptions about the distribution of unlabeled data. Our framework, grounded in a probabilistic model, innovatively refines the expectation-maximization algorithm. Our method showcases consistent state-of-the-art performance across diverse benchmarks and data distribution scenarios.
  arXiv  Detail & Related papers  (2024-02-21T03:39:04Z)
• Predictive World Models from Real-World Partial Observations [66.80340484148931]
  We present a framework for learning a probabilistic predictive world model for real-world road environments. While prior methods require complete states as ground truth for learning, we present a novel sequential training method to allow HVAEs to learn to predict complete states from partially observed states only.
  arXiv  Detail & Related papers  (2023-01-12T02:07:26Z)
• Learning and Predicting Multimodal Vehicle Action Distributions in a Unified Probabilistic Model Without Labels [26.303522885475406]
  We present a unified probabilistic model that learns a representative set of discrete vehicle actions and predicts the probability of each action given a particular scenario. Our model also enables us to estimate the distribution over continuous trajectories conditioned on a scenario, representing what each discrete action would look like if executed in that scenario.
  arXiv  Detail & Related papers  (2022-12-14T04:01:19Z)
• Sampling from Arbitrary Functions via PSD Models [55.41644538483948]
  We take a two-step approach by first modeling the probability distribution and then sampling from that model. We show that these models can approximate a large class of densities concisely using few evaluations, and present a simple algorithm to effectively sample from these models.
  arXiv  Detail & Related papers  (2021-10-20T12:25:22Z)
• Why do classifier accuracies show linear trends under distribution shift? [58.40438263312526]
  accuracies of models on one data distribution are approximately linear functions of the accuracies on another distribution. We assume the probability that two models agree in their predictions is higher than what we can infer from their accuracy levels alone. We show that a linear trend must occur when evaluating models on two distributions unless the size of the distribution shift is large.
  arXiv  Detail & Related papers  (2020-12-31T07:24:30Z)
• Characterizing and Avoiding Problematic Global Optima of Variational Autoencoders [28.36260646471421]
  Variational Auto-encoders (VAEs) are deep generative latent variable models. Recent work shows that traditional training methods tend to yield solutions that violate desiderata. We show that both issues stem from the fact that the global optima of the VAE training objective often correspond to undesirable solutions.
  arXiv  Detail & Related papers  (2020-03-17T15:14:25Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.