SolarWM: Open Data and Scalable Training for Long-Horizon Video World Models

• URL: http://arxiv.org/abs/2609.02886v1
• Date: Wed, 02 Sep 2026 17:59:41 GMT
• Title: SolarWM: Open Data and Scalable Training for Long-Horizon Video World Models
• Abstract summary: We introduce SolarWM, a fully open foundation for building interactive video world models from data preparation through long-horizon inference.<n>SolarWM addresses this coupling with a reconfigurable multi-source data engine and a backbone-native adaptation framework.<n>We instantiate four 5B--33B models based on Wan2.2, LTX-2.5, and MiniMax-H3 while preserving their native representations and objectives.<n>The resulting causal models enable real-time interaction over rollouts ranging from minutes to hours after being trained on only 5s sequences.
• Score: 88.64689505355055
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: We introduce SolarWM, a fully open foundation for building interactive video world models from data preparation through long-horizon inference. Training across heterogeneous data sources and video backbones is challenging: datasets differ in temporal scale, camera geometry, visual quality, motion, and captioning styles, while video generators use distinct representations and architectures. Naive data mixing and model-specific implementations therefore produce inconsistent supervision and make results difficult to reproduce and compare. SolarWM addresses this coupling with a reconfigurable multi-source data engine and a backbone-native adaptation framework. The engine converts 1.43 million canonical clips from 10 datasets into a unified, frame-aligned contract covering visual observations, metric camera geometry, captions, quality metadata, selection decisions, and provenance, while decoupling source processing from mixture construction. Under shared camera-conditioning, training, and inference interfaces, we instantiate four 5B--33B models based on Wan2.2, LTX-2.5, and MiniMax-H3 while preserving their native representations and objectives. A unified three-stage recipe combines bidirectional adaptation, teacher-forced autoregressive initialization, and distribution matching distillation. The resulting causal models enable real-time interaction over rollouts ranging from minutes to hours after being trained on only 5s sequences. By releasing the resulting data, pipeline, recipes, weights, and framework, SolarWM provides a reproducible and extensible foundation for interactive world-model research.

Related papers

• EchoWM: Open and Enterable Omnimodal World Models [71.49148067556887]
  EchoWM is an omnimodal world model for enterable generative media.<n>It responds to continuous navigation while jointly generating 720p video, environmental sound, music and speech.
  arXiv  Detail & Related papers  (2026-08-24T12:39:59Z)
• Matrix-Game 3.0: Real-Time and Streaming Interactive World Model with Long-Horizon Memory [53.39687409541093]
  Matrix-Game 3.0 is a memory-augmented interactive world model designed for 720p real-time longform video generation.<n>We introduce systematic improvements across data, model, and inference.<n> Experimental results show that Matrix-Game 3.0 achieves up to 40 FPS real-time generation at 720p resolution with a 5B model.
  arXiv  Detail & Related papers  (2026-04-10T06:00:09Z)
• Using Vision Language Foundation Models to Generate Plant Simulation Configurations via In-Context Learning [6.254251081017878]
  This paper introduces a benchmark to evaluate the performance of vision language models (VLMs) in generating plant simulation for digital twins.<n>We propose a novel approach that leverages state-of-the-art open-source VLMs to directly generate simulation parameters in format from drone-based remote sensing images.
  arXiv  Detail & Related papers  (2026-03-09T20:58:43Z)
• Latent Particle World Models: Self-supervised Object-centric Stochastic Dynamics Modeling [51.40150411616207]
  We introduce Latent Particle World Model (LPWM), a self-supervised object-centric world model scaled to real-world multi-object datasets.<n>LPWM autonomously discovers keypoints, bounding boxes, and object masks directly from video data.<n>Our architecture is trained end-to-end purely from videos and supports flexible conditioning on actions, language, and image goals.
  arXiv  Detail & Related papers  (2026-03-04T19:36:08Z)
• Disentanglement Beyond Static vs. Dynamic: A Benchmark and Evaluation Framework for Multi-Factor Sequential Representations [14.972702558607557]
  We introduce the first standardized benchmark for evaluating multi-factor sequential disentanglement across six diverse datasets.<n>We propose a post-hoc Latent Exploration Stage to automatically align latent dimensions with semantic factors, and introduce a Koopman-inspired model that achieves state-of-the-art results.<n>Our code is available on GitHub, and the datasets and trained models are available on Hugging Face.
  arXiv  Detail & Related papers  (2025-10-20T08:58:23Z)
• MORPH: Shape-agnostic PDE Foundation Models [37.26306668589026]
  MORPH is a shape-agnostic, autoregressive foundation model for partial differential equations (PDEs)<n>It is built on a convolutional vision backbone that seamlessly handles heterogeneous evaluations of varying data dimensionality (1D--3D)<n>Across extensive datasets, MORPH matches or surpasses strong baselines and recent state-of-the-art models.
  arXiv  Detail & Related papers  (2025-09-25T22:38:36Z)
• Matrix-Game 2.0: An Open-Source, Real-Time, and Streaming Interactive World Model [15.16063778402193]
  Matrix-Game 2.0 is an interactive world model generates long videos on-the-fly via few-step auto-regressive diffusion.<n>It can generate high-quality minute-level videos across diverse scenes at an ultra-fast speed of 25 FPS.
  arXiv  Detail & Related papers  (2025-08-18T15:28:53Z)
• HuggingGraph: Understanding the Supply Chain of LLM Ecosystem [6.131279654327215]
  Large language models (LLMs) leverage deep learning architectures to process and predict sequences of words.<n>LLMs can inherit vulnerabilities, biases, or malicious components that exist in previous models or datasets.<n>Motivated by that, this project aims to study such relationships between models and datasets.
  arXiv  Detail & Related papers  (2025-07-17T17:34:13Z)
• Genesis: Multimodal Driving Scene Generation with Spatio-Temporal and Cross-Modal Consistency [32.16082566679126]
  We present a unified framework for joint generation of driving videos and LiDAR sequences.<n>We employ a two-stage architecture that integrates a DiT-based video diffusion model with 3D-VAE modalities, and a BEV-aware LiDAR generator with NeRF-based rendering and adaptive sampling.<n>To guide the generation with structured semantics, we introduce DataCrafter, a captioning module built on vision-level models that provides scene-level and instance-language supervision.
  arXiv  Detail & Related papers  (2025-06-09T07:20:49Z)
• VideoMolmo: Spatio-Temporal Grounding Meets Pointing [66.19964563104385]
  VideoMolmo is a model tailored for fine-grained pointing of video sequences.<n>A novel temporal mask fusion employs SAM2 for bidirectional point propagation.<n>To evaluate the generalization of VideoMolmo, we introduce VPoMolS-temporal, a challenging out-of-distribution benchmark spanning five real-world scenarios.
  arXiv  Detail & Related papers  (2025-06-05T17:59:29Z)
• Data-Juicer 2.0: Cloud-Scale Adaptive Data Processing for and with Foundation Models [83.65386456026441]
  Data-Juicer 2.0 is a data processing system backed by 100+ data processing operators spanning text, image, video, and audio modalities.<n>It supports more critical tasks including data analysis, synthesis, annotation, and foundation model post-training.<n>The system is publicly available and has been widely adopted in diverse research fields and real-world products such as Alibaba Cloud PAI.
  arXiv  Detail & Related papers  (2024-12-23T08:29:57Z)

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.