EvolvR: Self-Evolving Pairwise Reasoning for Story Evaluation to Enhance Generation

Researchers have developed EvolvR, a self-evolving framework that improves AI's ability to evaluate and generate stories through pairwise reasoning and multi-agent data filtering. The system achieves state-of-the-art performance on three evaluation benchmarks and significantly enhances story generation quality when used as a reward model.

• →EvolvR framework addresses limitations in current AI story evaluation methods by combining self-synthesis and multi-agent filtering.
• →The system uses pairwise comparison and Chain-of-Thought reasoning to improve evaluation accuracy.
• →Achieved state-of-the-art performance on StoryER, HANNA, and OpenMEVA benchmarks.
• →When deployed as a reward model, it significantly improves the quality of AI-generated stories.
• →The approach bridges the gap between prompt engineering limitations and fine-tuning deficiencies in story evaluation.