Abstract
The Patent-Based Idea Generation task asks systems to turn real patents into product ideas viable within three years. We propose MK2, a prompt-centric pipeline: Gemini 2.5 drafts and iteratively edits a prompt, grafting useful fragments from weaker outputs; GPT-4.1 then uses this prompt to create one idea per patent, and an Elo loop judged by Qwen3-8B selects the best prompt-all without extra training data. Across three domains, two evaluator types, and six criteria, MK2 topped the automatic leaderboard and won 25 of 36 tests. Only the materials-chemistry track lagged, indicating the need for deeper domain grounding; yet, the results show that lightweight prompt engineering has already delivered competitive, commercially relevant ideation from patents.
Abstract (translated)
基于专利的想法生成任务要求系统将真实的专利转化为三年内可行的产品创意。我们提出了MK2，这是一个以提示为中心的流程：Gemini 2.5 负责起草和迭代编辑一个提示，并从较弱的输出中嫁接有用的片段；随后，GPT-4.1 使用此提示为每项专利生成一个想法；最后通过由Qwen3-8B评判的Elo循环来选择最佳提示——整个过程无需额外训练数据。在三个领域、两种评估类型和六项标准下，MK2 在自动排行榜上名列前茅，并赢得了36个测试中的25个。只有材料化学赛道略显落后，这表明需要更深入的领域知识；然而，结果已经显示，轻量级提示工程已经能够从专利中生成有竞争力且具有商业相关性的创意想法。
URL
https://arxiv.org/abs/2507.08335
