Source: arXiv · cs.CLView original ↗
Copyright remains with the original source. This site only collects, translates, or reformats the material.
What happened
arXiv:2608.12336v1 Announce Type: new Abstract: A story premise is the creative spark from which a full narrative can grow. Yet LLM-based story generation has mostly emphasized later-stage planning, controllability, coh
Analysis and impact
导读摘要
正文
故事前提(story premise)是叙事从无到有的第一颗创意种子,但现有的大语言模型故事生成研究却大多绕开了这一步。这篇来自 arXiv 的论文提出,目前学界与业界的注意力主要集中在故事生成的后端环节——例如情节规划、叙事可控性以及连贯性维护——而故事前提的生成质量本身,反而缺乏系统性的探索。
针对这一空白,论文提出了一种名为 StorySpark 的方法,其核心思路是「模块化进化搜索」。从摘要提供的信息来看,该方法将故事前提的生成过程拆解为若干可独立演化的模块,并通过进化搜索的方式在候选空间中迭代筛选,以寻找更具创意潜力、更能支撑后续叙事扩展的前提。摘要原文在关键处截断,关于模块的具体划分方式、进化算法的适应度函数设计以及实验所用的评测基准,原文未提供完整细节。
将这篇工作放回同日发布的 arXiv 论文群中,可以看到一个共同的研究关切:大语言模型在「知道规则」与「实际使用规则」之间存在系统性落差。例如,同日发布的另一篇论文讨论了 LLM 在语用约束推理中的激活瓶颈,指出模型即使掌握了隐含的可行性约束,也会在显著表面线索的干扰下做出错误判断。StorySpark 所针对的问题——模型能生成文本,却未必能产出真正具有叙事张力的前提——与这一观察在逻辑上相通:生成能力本身并不自动等同于创意质量。
从方法层面看,进化搜索在文本生成领域的应用并非全新思路,但将其专门用于故事前提这一早期创意阶段,并以模块化方式组织搜索空间,构成了该工作的主要区分点。论文的实际效果如何、相比直接让 LLM 一次性生成前提能带来多大提升,仍需等待完整论文或后续复现实验来验证。对于关注计算创意与叙事智能的研究者而言,这篇论文提供了一个值得追踪的方向:它把「创意起点」本身当成了一个需要专门优化的技术问题,而非默认 LLM 能够自然胜任的前置步骤。
References
Original source text
arXiv:2608.12336v1 Announce Type: new Abstract: A story premise is the creative spark from which a full narrative can grow. Yet LLM-based story generation has mostly emphasized later-stage planning, controllability, coherence, and prose expansion, while premise-level ideation remains comparatively underexplored. We introduce StorySpark, a module-wise evolutionary search framework for story premise generation. StorySpark operates over interpretable narrative modules such as background, persona, event, ending, and twist, treating each active module not as a static field to fill once, but as a local search space conditioned on the partial premise built so far. For each module, it generates alternatives, evaluates them in context, refines them through feedback-driven mutation and recombination, preserves complementary strengths with Pareto-guided selection, and reallocates frontier capacity to balance branch coverage with promising directions. Multi-view automatic and human evaluations show that StorySpark produces stronger final premises than competitive baselines, with especially consistent gains in originality; when expanded with the same story writer, its premises also lead to higher-quality downstream stories while maintaining completeness, fascination, and diverse usable narrative directions.