🤖 AI Summary
This work addresses the tendency of current large language models to suffer from semantic convergence when generating scientific ideas, which undermines both novelty and diversity. To mitigate premature convergence, the study introduces an evolutionary computation framework—the first of its kind applied to this task—featuring a population-based search strategy. The proposed approach incorporates rank-driven mutation, semantic-aware crossover, differentiated retrieval planning, and a lightweight evaluation mechanism. Experimental results demonstrate that the method significantly outperforms baseline approaches, improving automatically assessed novelty from 0.1 to 0.4 and diversity from 0.24 to 0.55, while maintaining high idea quality.
📝 Abstract
Generating novel research ideas is fundamental to scientific progress. While Large Language Models (LLMs) show promise in assisting this process, existing approaches often exhibit semantic convergence, resulting in limited diversity and novelty. To address this, we introduce EvoGens, an evolution-inspired framework that recasts scientific idea generation as an evolutionary search over a population of ideas. EvoGens iteratively applies rank-based mutation with differentiated retrieval planning to incorporate external knowledge, and semantic-aware crossover to fuse complementary concepts for conceptual reorganization. A lightweight evaluation signal guides the selection process, encouraging sustained exploration while mitigating premature convergence. Extensive experiments demonstrate that EvoGens substantially enhances exploration capabilities compared to state-of-the-art baselines. Specifically, it improves the Novelty from 0.1 to 0.4 and the Diversity from 0.24 to 0.55, while maintaining comparable idea quality under the current automatic evaluation protocol. These findings suggest that evolutionary mechanisms can serve as a useful framework for exploration-oriented research ideation, especially for broadening the novelty and diversity of candidate ideas under a shared automatic evaluation setting.