{"paper":{"title":"UI2Code^N: UI-to-Code Generation as Interactive Visual Optimization","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"UI-to-code generation improves by treating it as a closed-loop visual optimization process rather than single-pass output.","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Jiale Cheng, Jie Tang, Mingde Xu, Weihan Wang, Wenyi Hong, Xiaotao Gu, Xinyue Fan, Zhen Yang","submitted_at":"2025-11-11T13:00:09Z","abstract_excerpt":"UI-to-code aims to translate UI screenshots into executable front-end code. Despite progress with vision-language models (VLMs), most existing methods formulate UI-to-code as a single-pass generation, which mismatches real-world UI development that is inherently iterative and feedback-driven. We reformulate UI-to-code as an interactive visual optimization problem, where code generation is embedded in a closed-loop process of execution, visual inspection, and iterative refinement driven by rendered visual feedback. To address the non-differentiability of visual objectives and the noise of absol"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"Experiments demonstrate state-of-the-art performance on UI drafting, UI polishing, and UI editing benchmarks, even outperforming larger models, with performance consistently improving through iterative visual optimization.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That rendered visual feedback can reliably drive iterative refinement despite non-differentiability of visual objectives and noise in absolute evaluators, which RVPO is claimed to address via relative rankings.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"UI2Code^N turns UI-to-code into a closed-loop visual optimization process using Relative Visual Policy Optimization on a 9B model, achieving SOTA results that improve with iterations.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"UI-to-code generation improves by treating it as a closed-loop visual optimization process rather than single-pass output.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"df71cf665818b0354a76b69c10fd51da32223d4911af141200c80cf0cbd33f9c"},"source":{"id":"2511.08195","kind":"arxiv","version":4},"verdict":{"id":"5b955968-47fb-427e-ac63-28ad1ca3fb6e","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-17T23:52:52.575301Z","strongest_claim":"Experiments demonstrate state-of-the-art performance on UI drafting, UI polishing, and UI editing benchmarks, even outperforming larger models, with performance consistently improving through iterative visual optimization.","one_line_summary":"UI2Code^N turns UI-to-code into a closed-loop visual optimization process using Relative Visual Policy Optimization on a 9B model, achieving SOTA results that improve with iterations.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That rendered visual feedback can reliably drive iterative refinement despite non-differentiability of visual objectives and noise in absolute evaluators, which RVPO is claimed to address via relative rankings.","pith_extraction_headline":"UI-to-code generation improves by treating it as a closed-loop visual optimization process rather than single-pass output."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2511.08195/integrity.json","findings":[],"available":true,"detectors_run":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938"},"references":{"count":0,"sample":[],"resolved_work":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","internal_anchors":0},"formal_canon":{"evidence_count":2,"snapshot_sha256":"505429aacd6062132bf8add998eaec6b85c0cc0fa3ea5ff43ca59c33a3a87f74"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"}