Searcharxiv⌕ Search

arXiv subjects

Xiangyu Pan

Publications and source records attributed to Xiangyu Pan.

2 recordsLinked to original sources

JointGV: Structure-Aware Correction of Visual World Models with Parallel Graph Dynamics for Long-Horizon Modeling

Visual world models predict future visual states conditioned on actions, but recursive errors in entity poses and spatial relations limit their long-horizon prediction for robot planning and control. We ask whether a separately predicted relational state can correct this drift without restructuring the visual model. This is challenging because graph nodes and visual states share neither elements nor a feature space. We propose Joint Graph--Visual Dynamics Correction (JointGV), a plug-in based on a cross-granularity dynamics hypothesis: despite their different representations, graph-node and visual-state changes describe the same physical transition. JointGV rolls out a task-conditioned scene graph alongside the visual world model, couples their predicted changes, and combines the resulting graph-motion context with graph-predicted node positions to compute a bounded residual. Graph-learning and correction gradients do not propagate into the original visual dynamics. We attach JointGV to DINO-WM (spatial tokens) and DreamerV3 (a global latent state), evaluating both on DMC Cheetah Run and Walker Walk across three seeds and extending the evaluation to contact-rich manipulation with DMC Finger Turn Hard. Across 32-step rollouts, JointGV improves structural and visual-prediction accuracy, reducing joint-position error by 11--28\%. For spatial tokens, learned graph--visual change coupling gives lower error than broadcasting the same mean graph-motion context to every token, particularly when non-adjacent parts appear close together in the image. On real-robot inverted-pendulum sequences, JointGV reduces DINO-WM's 32-step pole-angle error from $12.56^\circ$ to $6.69^\circ$ and improves DreamerV3 and LeWorldModel as well. In simulated inverted-pendulum control, an actor trained with JointGV-corrected rollouts completes 46 of 50 episodes, compared with none using the original rollouts.

cs.RO↗

Lift-independence problem in the $p$-adic Simpson correspondence for curves

Let $X$ be a proper smooth rigid analytic variety over a complete algebraically closed $p$-adic field $\mathbf C$. Fix a continuation $\operatorname{Exp}$ of $\exp$. Faltings, in the curve case, and Heuer showed that any lifting $\widetilde X$ of $X$ over $B_{\mathrm{dR}}^+/t^2$ induces an equivalence between the category of Higgs bundles on $X_{\text{ét}}$ and the category of $v$-vector bundles on $X_v$. Changing $\widetilde X$ is known to alter the correspondence nontrivially. This raises the natural question of how the isomorphism class of the Higgs bundle associated with a fixed $v$-vector bundle varies with $\widetilde X$, a question that has not been systematically studied. In this paper, we address this question when $X$ is a curve of genus $g\ge 2$. More precisely, we call a Higgs bundle lift-independent if it corresponds to the same $v$-bundle under the $p$-adic Simpson correspondence for every lifting $\widetilde X$. We prove the following results. (1) Every lift-independent Higgs bundle of rank $r\le g$ is nilpotent. (2) Every semistable lift-independent Higgs bundle of rank $r\le \sqrt{g}+1$ has zero Higgs field. (3) In contrast, vanishing fails even within the class singled out by Faltings' conjecture: there always exists a lift-independent semistable Higgs bundle of degree $0$ with nonzero Higgs field.

math.AG↗