FailureAtlas：通过主动探索绘制T2I模型的失败景观
FailureAtlas: Mapping the Failure Landscape of T2I Models via Active Exploration
提出主动探索范式替代静态基准评测，通过结构化搜索输入空间发现T2I模型的系统性失败模式，并首次大规模关联训练数据稀缺性，为生成模型鲁棒性评估提供可复用引擎。
原文依据 FailureAtlas performs guided exploration in the input space, framing error discovery as a structured search for minimal, failure-inducing concepts.
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论静态基准驱动的评估为分析文本到图像（T2I）模型提供了宝贵基础。然而，基准中固定且预定义的提示词集本质上限制了诊断深度，难以揭示模型系统性失败的全貌或分离其根本原因。我们主张一种互补范式：主动探索，并引入FailureAtlas——首个旨在大规模自主探索和绘制T2I模型广阔失败景观的框架。与评估固定提示词集的基准不同，FailureAtlas在输入空间中进行引导式探索，将错误发现构建为对最小失败诱导概念的结构化搜索。尽管这是一个计算爆炸性问题，我们通过新颖的加速技术使其变得可行。应用于Stable Diffusion模型时，我们的方法发现了数十万个先前未知的错误切片（例如，仅在SD1.5中超过247,000个），并首次提供大规模证据将这些失败与训练集中的数据稀缺性联系起来。通过为深度模型审计提供原则性且可扩展的引擎，FailureAtlas建立了一种新的、诊断优先的方法论，以指导更鲁棒的生成式AI的开发。代码可在https://github.com/cure-lab/FailureAtlas获取。
英文原文摘要 1338 字符
Static benchmark-driven evaluation has provided a valuable foundation for analyzing Text-to-Image (T2I) models.However, the fixed and predetermined prompt sets in benchmarks inherently limit diagnostic depth, making it difficult to uncover the full landscape of models' systematic failures or isolate their root causes.We argue for a complementary paradigm: active exploration, and introduce FailureAtlas, the first framework designed to autonomously explore and map the vast failure landscapes of T2I models at scale.Unlike benchmarks that evaluate a fixed prompt set, FailureAtlas performs guided exploration in the input space, framing error discovery as a structured search for minimal, failure-inducing concepts. While this is a computationally explosive problem, we make it tractable with novel acceleration techniques. When applied to Stable Diffusion models, our method uncovers hundreds of thousands of previously unknown error slices (e.g., over 247,000 in SD1.5 alone) and provides the first large-scale evidence linking these failures to data scarcity in the training set. By providing a principled and scalable engine for deep model auditing, FailureAtlas establishes a new, diagnostic-first methodology to guide the development of more robust generative AI. The code is available at https://github.com/cure-lab/FailureAtlas.