AutoPR: Let's Automate Your Academic Promotion [pdf] (arxiv.org)

Researchers introduced AutoPR, a new task and toolkit for automating the creation of high-quality promotional content for research papers to improve discoverability and impact. To evaluate systems, they released PRBench, a multimodal benchmark that pairs 512 peer-reviewed papers with human-quality promotional posts and scores systems on Fidelity (accuracy and tone), Engagement (appeal and audience targeting), and Alignment (timing and channel optimization). The paper also stresses the growing need for scalable scholarly communication as publication volume explodes and authors struggle to reach audiences. To tackle AutoPR the authors propose PRAgent, a multi-agent pipeline with three stages: multimodal content extraction (text, figures, metadata), collaborative synthesis to craft polished copy, and platform-specific adaptation that optimizes tone, tags, and timing for each channel. On PRBench, PRAgent markedly outperforms direct LLM baselines—delivering a 604% increase in total watch time, 438% more likes, and at least a 2.9× boost in overall engagement. Ablations show platform modeling and targeted promotion yield the largest gains. The work formalizes automated scholarly promotion as a measurable problem, supplies data and demos, and highlights practical gains and risks (e.g., fidelity vs. virality trade-offs) for downstream adoption in research communication pipelines.