Odysseus Navigates the Sirens' Song: Dynamic Focus Decoding for Factual and Diverse Open-Ended Text Generation

Large Language Models (LLMs) are increasingly required to generate text that is both factually accurate and diverse across various openended applications. However, current stochastic decoding methods struggle to balance such objectives. We introduce Dynamic Focus Decoding (DFD), a novel plug-and-play stochastic approach that resolves this trade-off without requiring additional data, knowledge, or models. DFD adaptively adjusts the decoding focus based on distributional differences across layers, leveraging the modular and hierarchical nature of factual knowledge within LLMs. This dynamic adjustment improves factuality in knowledge-intensive decoding steps and promotes diversity in less knowledge-reliant steps. DFD can be easily integrated with existing decoding methods, enhancing both factuality and diversity with minimal computational overhead. Extensive experiments across seven datasets demonstrate that DFD significantly improves performance, providing a scalable and efficient solution for open-ended text generation. 1 * Corresponding author 1 Code is publicly available at https://github.com/ lllllw-222/Siren-DFD Who formulated the laws of motion? Fixed High Temperature r1: Isaac Newton was the one who formulated the laws of motion. r2: Sir Isaac Newton, who was born on November 19, 1643 in England. r3: Galileo Galilei formulated the laws of motion.