Any2Poster: Any-Source Poster Generation Across Modalities and Domains

AI can produce polished scientific artifacts faster than those artifacts can be verified. We study posters as a stress case: they compress dense sources into a single visual page, yet existing evaluations are often limited to papers or surface-level visual similarity. We introduce Any2Poster Bench, a benchmark for any-source poster generation across eight input modalities—PDFs, URLs, PPTX, DOCX, Markdown, LaTeX, notebooks, and videos—and five domains. Its BenchQuiz protocol tests verbatim and interpretive information recovery, complemented by VLM judgments of visual communication and validated against human readers. We further present Any2Poster Agent, an end-to-end reference agent that parses heterogeneous sources, plans and renders editable posters, and iteratively repairs them using visual feedback. On Any2Poster Bench, Any2Poster Agent achieves 87.25% average accuracy across modalities and 87.28% across domains. On PaperQuiz-style evaluation, it improves over PosterAgent-4o from 51.06–51.33% to 72.58% overall accuracy and from 116–121 to 145.16 in density-augmented score. Together, Any2Poster Bench and Any2Poster Agent provide source-grounded verification infrastructure and a competitive baseline for multimodal, domain-general poster generation.