Physics-based urban simulations require editable, topology-valid building representations, yet urban data are typically available only as images or segmentation masks. We propose Universal Anchor Representation (UAR), a normalized hierarchical parameterization that constrains inference to bounded coordinates (t,s) \in [0,1]2 while guaranteeing structural connectivity by construction. We pair UAR with a domain-specific language and compiler to produce executable parametric procedural graphs, and develop a vision-language model (VLM) guided pipeline with overlay-based refinement to recover them from top-view building images. Experiments on 110 real urban buildings spanning 11 morphological categories demonstrate that UAR-based generation achieves 75.2% F1 score and 100% compilation success, substantially outperforming absolute-coordinate generation, which attains 43.1% F1 and 57% success under identical model and prompt configurations. Wind simulation case studies further confirm that recovered parameters yield physically interpretable changes in flow fields under parametric edits.
Jaeyeon Kim, Po-Yen Lai, J. Wong et al.· Proceedings of the 32nd ACM...· 0 citations
Long-horizon web agents frequently fail without knowing where or why execution broke down. This issue is particularly pronounced in skill-based agentic web systems, where failures arise within composite skills whose internal decision processes are not directly traceable, making precise diagnosis and repair especially difficult over long horizons. We introduce SkillTracer, a framework that represents skills as attributed plan graphs structured by hierarchical nodes and verifiable edge transitions, enabling programmatic verification of execution progress. By decomposing skills into inspectable hierarchies, SkillTracer converts raw interaction traces into structural evidence, making execution breakdowns localizable to specific node-level decision points and attributable to failing components. This attribution signal facilitates targeted structural repair, allowing the agent to selectively revise failing components while preserving the integrity of valid substructures for partial reuse and adaptive recovery. Furthermore, SkillTracer synthesizes short-term traces with long-term historical evidence to construct a persistent skill graph, enabling failure patterns to drive continual refinement across episodes. Evaluated on challenging long-horizon benchmarks, SkillTracer achieves a 17.7% average improvement in success rate over strong baselines, with gains of up to 56.3% in cross-domain settings, demonstrating that structural attribution and skill repair are critical for reliable long-horizon web interaction. A project page is available at: https://liyuuuuy.github.io/SkillTracer/.
Yuyang Li, Yiran Dou, Jiejing Shao et al.· Proceedings of the 32nd ACM...· 5 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.
