Paper 2603.05890
Lost in Stories: Consistency Bugs in Long Story Generation by LLMs
- Published
- Mar 2026
- Research lab
- Independent
- Citations
- 3
- GitHub
- 192 stars
01 In brief
Summary
The paper introduces ConStory-Bench, a benchmark for evaluating narrative consistency in long-form story generation by LLMs, and CONSTORY-CHECKER, an automated pipeline that detects contradictions with textual evidence.
The benchmark includes 2,000 prompts across four task scenarios and a taxonomy of five error categories with 19 subtypes.
Evaluation of various LLMs reveals that consistency errors are most common in factual and temporal dimensions, tend to appear around the middle of narratives, occur in segments with higher token-level entropy, and certain error types co-occur.
GPT-5-Reasoning achieved the lowest error density (CED 0.113), while most models still struggle with long-form consistency.
The study also finds that errors increase linearly with output length, and that entropy can serve as an early warning signal for potential inconsistencies.
02 From the paper
Abstract
What happens when a storyteller forgets its own story? Large Language Models (LLMs) can now generate narratives spanning tens of thousands of words, but they often fail to maintain consistency throughout. When generating long-form narratives, these models can contradict their own established facts, character traits, and world rules. Existing story generation benchmarks focus mainly on plot quality and fluency, leaving consistency errors largely unexplored. To address this gap, we present ConStory-Bench, a benchmark designed to evaluate narrative consistency in long-form story generation. It contains 2,000 prompts across four task scenarios and defines a taxonomy of five error categories with 19 fine-grained subtypes. We also develop ConStory-Checker, an automated pipeline that detects contradictions and grounds each judgment in explicit textual evidence. Evaluating a range of LLMs through five research questions, we find that consistency errors show clear tendencies: they are most common in factual and temporal dimensions, tend to appear around the middle of narratives, occur in text segments with higher token-level entropy, and certain error types tend to co-occur. These findings can inform future efforts to improve consistency in long-form narrative generation. Our project page is available at https://picrew.github.io/constory-bench.github.io/.