arXiv：cs.AI· Kunwei Wu, Xiang Liu, Guocai Yao, Junming Chen, Zhikang Chen, Min Zhang, Pengwei Wang, Sen Cui·· 4 小时前AI 评分34
LPA-CWM：用学习式物理裁决器与反事实世界模型做运动推理
LPA-CWM: A Learned Physical Adjudicator for Motion Reasoning with Counterfactual World Models
AI 导读
LPA-CWM 提出轻量级学习式物理裁决器（LPA），在 CWM 冻结的前提下为无序候选集预测相对权重，以融入物理先验。该 LPA 仅 3.0M 参数，在 MOVi-F 轨迹上训练，并配合窗口化定位与一次配对重评估恢复运动。
正文
Abstract:Counterfactual world models (CWM) extract motion from pretrained video predictors by comparing factual and intervened predictions, but uniform aggregation weights responses equally without explicitly incorporating physical priors. Our key insight is to incorporate physical priors into candidate reliability learning, motivating LPA-CWM with a lightweight Learned Physical Adjudicator (LPA). Trained on dense MOVi-F trajectories, the 3.0M-parameter LPA compares visual context and response structure across an unordered candidate set to predict relative weights; windowed localization and one paired re-evaluation recover motion with the CWM frozen. Existing video-level benchmarks do not directly assess motion correspondence, where low localization error can conceal missing trajectory segments. We introduce Completeness-aware Motion Correspondence (CMC), a ground-truth-anchored protocol jointly measuring localization, completeness, visibility, and continuity, counting missing predictions as failures on visible dynamic points. Across DAVIS, Kinetics, and RoboTAP, LPA-CWM improves all main CMC measures over Uniform CWM, with relative gains of 18.1%--60.0% in average Dynamic Correspondence Accuracy ($\mathrm{DCA}_{\mathrm{avg}}$), and improves TAP-Vid First tracking accuracy (overview: this https URL).
| Comments: | A quick overview is available at this https URL |
| Subjects: | Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Robotics (cs.RO) |
| Cite as: | arXiv:2609.14073 [cs.CV] |
| (or arXiv:2609.14073v2 [cs.CV] for this version) | |
| https://doi.org/10.48550/arXiv.2609.14073 arXiv-issued DOI via DataCite |
Submission history
From: Xiang Liu [view email]
[v1]
Sat, 12 Sep 2026 17:46:32 UTC (2,355 KB)
[v2]
Fri, 2 Oct 2026 05:10:28 UTC (4,588 KB)
来源：arXiv：cs.AI · arxiv.org