发表机构
Pfizer; University of California San Diego; Institute for Medical Philosophy and Future Artificial Intelligence; Tongji University; Shanghai Jiao Tong University; Institute of Humanities and Social Science Data, Fudan University（辉瑞公司; 加利福尼亚大学圣地亚哥分校; 医学哲学与未来人工智能研究所; 同济大学; 上海交通大学; 复旦大学人文社会科学数据研究所）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
研究大语言模型递归自精炼工作流程的长期动态。通过生成精炼轨迹并分析多种指标，发现轨迹迅速饱和，编辑多在前几次迭代，确定性解码表现更好，平均编辑幅度呈指数松弛，收敛摘要有多种优势，支持动态系统观点并推动停止标准。
AI 中文摘要
大语言模型越来越多地用于递归精炼工作流程，即由同一模型反复修订初始草稿。尽管其应用日益广泛，但此类工作流程的长期动态仍知之甚少。反复精炼会无限期地持续改进输出，还是会趋向于稳定的文本形式？我们将递归自精炼视为一个动态过程，其中LLM的反复修订将文本推向模型偏好的软定点区域。使用GPT-5.5，我们在默认温度和确定性解码下为50篇ICML 2025摘要生成了10步精炼轨迹，并额外评估了15篇ICML 2020摘要。我们分析了归一化编辑距离、精确和近似定点、字数稳定性、指数松弛以及外部LLM作为评判的评估。在所有设置中，精炼轨迹迅速饱和。大多数编辑发生在前几次迭代中，之后轨迹进入软定点区域，只有微小的表面变化。确定性解码比默认温度解码更早达到精确定点，且残余波动更小，两者都实现了普遍近似收敛。平均编辑幅度遵循一致的指数松弛模式，表明趋向于模型偏好的文本平衡而非开放式优化。外部评估表明，收敛后的摘要在保留技术含义的同时提高了清晰度、简洁性和科学风格。这些发现支持了LLM自精炼的动态系统观点，并推动了基于编辑幅度饱和的实际停止标准。
英文摘要
Large language models are increasingly used in recursive refinement workflows, where an initial draft is repeatedly revised by the same model. Despite their growing use, the long-term dynamics of such workflows remain poorly understood. Does repeated refinement continue to improve outputs indefinitely, or does it converge toward a stable textual form? We study recursive self-refinement as a dynamical process in which repeated LLM revision drives text toward a model-preferred soft fixed-point region. Using GPT-5.5, we generate 10-step refinement trajectories for 50 ICML 2025 abstracts under both default-temperature and deterministic decoding, and additionally evaluate 15 ICML 2020 abstracts. We analyze normalized edit distance, exact and approximate fixed points, word-count stability, exponential relaxation, and external LLM-as-a-judge evaluation. Across all settings, refinement trajectories rapidly saturate. Most edits occur within the first few iterations, after which trajectories enter a soft fixed-point region with only minor surface-level changes. Deterministic decoding reaches exact fixed points earlier and exhibits smaller residual fluctuations than default-temperature decoding, while both achieve universal approximate convergence. The average edit magnitude follows a consistent exponential relaxation pattern, suggesting convergence toward a model-preferred textual equilibrium rather than open-ended optimization. External evaluation indicates that converged abstracts improve clarity, conciseness, and scientific style while preserving technical meaning. These findings support a dynamical-systems view of LLM self-refinement and motivate practical stopping criteria based on edit-magnitude saturation.
