Video World Models · 2026-11-10
QCC-WM: Linking Macro-State Dynamics and Visual Generation via Quotient Causal Construction
Zhizhong Fu, Wei Zhou, Zhaoyang Jiang, Yudong Lin, Songlun Li, Zhi Li, Feifan Chen, Gaoming Li, Xiaorong Ding, Yifan Chen
AI-assisted summary based on the paper. Consult the linked source for methods and limitations. Benchmark numbers appear after editor verification.
TL;DR
QCC-WM is an action-conditioned visual world model that constructs macro-states as quotient-based causal macro-nodes. Action responses guide macro-variable construction and a shared reference quotient measurement constrains consistency between predicted macro-states and observed changes. Macro predictions drive motion-guided reference transport and readback with node-level error correction for future visual generation. On US-VLA, it reports improved ROI-MSE/LPIPS for multi-step ultrasound video prediction.
Why it matters
The authors address a core challenge in action-conditioned world modeling: forming macro-states from high-dimensional observations so recursive action-conditioned predictions remain consistent with visual evolution. QCC-WM links macro-state dynamics to image generation using a fixed reference quotient measurement, motion-guided reference transport, and spatially mapped node-level error feedback.
Method
- Constructs macro-nodes using action-response-guided soft quotient mappings, aggregating local visual features into spatially corresponding macro states.
- Per prediction window, fixes a reference quotient measurement (shared assignment/content alignment) to define consistent macro-state evolution, supervising states and temporal increments.
- Generates future observations by transporting reference appearance via predicted motion, then performs readback under the same quotient measurement and maps remaining node-level errors back to spatial regions for correction.
Limitation
Overall, QCC maintains favorable perceptual quality during multi-step extrapolation, while structural similarity at longer horizons remains an area for improvement.
Abstract (from arXiv)
A key challenge in action-conditioned world modeling is to construct macro-states from high-dimensional observations that support recursive prediction under actions, while keeping their predicted evolution consistent with visual changes in generated observations. We propose QCC-WM, a visual world model based on Quotient-based Causal Construction. QCC-WM uses action responses to guide macro-variable construction for recursive action-conditioned prediction. A shared reference quotient measurement constrains the consistency between predicted states and observed changes, allowing macro-level predictions to provide structured constraints for future visual generation. Using translational and rotational actions and their predicted responses, the model aggregates local visual features into macro-nodes with spatial correspondence through causal abstraction, and recursively predicts node content and motion under a shared reference measurement. Macro-level predictions guide future observation generation, while predicted motion transports historical appearance information. Node-level errors between generated observations and macro-level targets are then mapped back to the corresponding spatial regions for correction, improving consistency between visual generation and macro-level predictions. On the US-VLA dataset, QCC-WM achieves an ROI-MSE of 0.01861 for four-step prediction. Compared with the best baseline at each prediction horizon, it reduces LPIPS by 48.59%, 41.68%, and 30.23% for four-step prediction and eight- and sixteen-step recursive extrapolation, respectively. Ablation studies and mechanism analyses show that motion-guided reference transport mainly improves pixel reconstruction, while node-level error feedback further improves perceptual quality and consistency between generated observations and macro-level predictions.
Related papers
- WorldCast: Exploring Intelligent Visual Generation in World State Prediction
- WorldFoundry: Towards Unified Infrastructure for World Intelligence
- V-JEPA Policy: Building Effective World-Action Models on Predictive Visual Latents
- Pretrained Features as Teachers, Not States: Compact Latent World Models for Visual Planning
- World-T3: Test-Time Training as Persistent Memory for Real-Time Video World Models
Publication record
Bibliographic evidence describes where and when a record appeared. It does not validate scientific claims.
First announced by arXiv: 2026-11-10 · Indexed by Pub-AI: 2026-11-10 UTC
QCC-WM: Linking Macro-State Dynamics and Visual Generation via Quotient Causal Construction
First announced by arXiv: 2026-11-10 · Original record ↗
- arxiv
- 2611.12353
Explicit source versions (1)
- v1 · Date not supplied
- Preprint available
Source date: 2026-11-10 · First detected: 2026-11-10 UTC · Historical baseline
Source evidence (1)
- Official evidence ↗ · RSS guid=oai:arXiv.org:2611.12353v1; arxiv:announce_type=new; item pubDate (official metadata)
Abstract · arxiv
A key challenge in action-conditioned world modeling is to construct macro-states from high-dimensional observations that support recursive prediction under actions, while keeping their predicted evolution consistent with visual changes in generated observations. We propose QCC-WM, a visual world model based on Quotient-based Causal Construction. QCC-WM uses action responses to guide macro-variable construction for recursive action-conditioned prediction. A shared reference quotient measurement constrains the consistency between predicted states and observed changes, allowing macro-level predictions to provide structured constraints for future visual generation. Using translational and rotational actions and their predicted responses, the model aggregates local visual features into macro-nodes with spatial correspondence through causal abstraction, and recursively predicts node content and motion under a shared reference measurement. Macro-level predictions guide future observation generation, while predicted motion transports historical appearance information. Node-level errors between generated observations and macro-level targets are then mapped back to the corresponding spatial regions for correction, improving consistency between visual generation and macro-level predictions. On the US-VLA dataset, QCC-WM achieves an ROI-MSE of 0.01861 for four-step prediction. Compared with the best baseline at each prediction horizon, it reduces LPIPS by 48.59%, 41.68%, and 30.23% for four-step prediction and eight- and sixteen-step recursive extrapolation, respectively. Ablation studies and mechanism analyses show that motion-guided reference transport mainly improves pixel reconstruction, while node-level error feedback further improves perceptual quality and consistency between generated observations and macro-level predictions.
