这篇论文告诉你，用模型自己问自己生成的问答数据来训练模型，可能会引入隐蔽的偏见和指令注入风险。他们用一个简单方法就能把问题率从88%压到13%，值得做数据合成的同学看看。
语言模型越来越多地使用自生成的问答数据来微调或蒸馏。研究表明，生成阶段并非中性预处理，而是隐含策略：模型在提问时不会均匀扫描文档，覆盖很快饱和并集中在显著区域，不同提示也会聚焦相同片段。在回答阶段，模型倾向于服从文本中嵌入的指令性内容，且这种服从取决于指令的表面形式和意图。作者通过将每个问题固定到特定目标降低了选择偏差，并在回答前过滤指令性文本，将注入合规率从88%降至13%，同时保留几乎所有干净文本。
Self-Study Reconsidered: The Hidden Fragility of Learning from Self-Generated QA
Language models are increasingly taught from synthetic question--answer (QA) supervision: a model generates questions about a document, answers them from the same text, and the resulting pairs are used to fine-tune, distill, or compress knowledge into another model. We show that this generation step is not neutral preprocessing. It is an implicit policy that both selects which evidence becomes training signal and decides how that evidence is answered, and it is fragile at both stages. When choosing what to ask, generators do not scan a document uniformly. Coverage saturates early and concentrates on salient spans, diverse prompts converge on the same regions, and what looks question-worthy is driven by local presentation. As a result, salient artifacts such as poorly cleaned markup can hijack question generation across model families and scales. When answering, the model that produces the supervision tends to obey instruction-like passages embedded in the text. This compliance depends on the intent and surface form of the passage rather than its strictness, and is worst under task conflict, where larger models comply more often. These failure modes arise from choices made during QA generation, so they can be reduced without changing the training loop. Tying each question to a fixed target reduces biased selection, and filtering instruction-like spans before answering lowers mean injection compliance from $88\%$ to $13\%$ in our evaluation while retaining nearly all clean text.