MK2 at PBIG Competition: A Prompt Generation Solution

2025-07-11

AI Summary

This study addresses the challenge of efficiently transforming patents into product ideation. We propose MK2, a lightweight, training-free prompt evolution framework. Methodologically, Gemini 2.5 iteratively refines an initial prompt by integrating effective fragments from low-quality outputs; GPT-4.1 generates product concepts using the optimized prompt; and Qwen3-8B orchestrates an Elo-based tournament to evaluate and select the best-performing prompts. Our key contribution lies in establishing a cross-model collaborative and game-theoretic prompt evolution mechanism, substantially enhancing the practicality of prompt engineering for commercial ideation. Experiments span three technical domains and employ six automated evaluation dimensions. MK2 outperforms baselines in 25 out of 36 comparative trials and achieves top overall performance—slightly trailing only in materials chemistry. Results confirm the framework’s strong generalizability and real-world applicability.

Technology Category

• Natural Language Processing: Prompt Engineering / Prompting
• Search and Optimization: Sampling/Simulation-based Search
• Cognitive Modeling & Cognitive Systems: Computational Creativity

Application Category

• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems
• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

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.

Problem

• Generating viable product ideas from patents
• Optimizing prompt engineering without training data
• Evaluating idea quality across multiple domains

Innovation

• Gemini 2.5 drafts and edits prompts iteratively
• GPT-4.1 generates one idea per patent
• Elo loop selects best prompt via Qwen3-8B