发表机构
Cornell Tech（康奈尔理工学院）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
研究语言模型在特定格式要求下答案多样性变化，通过对44个模型进行31个类别提示实验，发现结构化输出使答案收敛、多样性降低，存在寄存器梯度，揭示了模型对不同格式的响应差异及对答案多样性的影响。
AI 中文摘要
当语言模型必须从大量同样有效的选项中选择一个答案时，一个格式子句——“仅以JSON格式回复”——会改变它选择的答案。我们重新运行了单字普查（arXiv:2607.12796）：向44个模型提出31个宽答案空间类别提示，现在要求以JSON格式回复——不进行模式强制，不进行约束解码，只有请求。收敛急剧加深：在无约束的“选一个词”提示中，模态答案从占库的41%升至64%，不同答案从52个降至36个；平均答案选择惊奇度从1.80比特降至1.58比特。这种影响是渐进的：44个模型中有6个单独变动（BH-FDR q =.10），都朝着模态方向，由最具特色的模型引领，而顺从底线不变。它是一个锐化器，而非重新索引器——31个类别中有28个类别中的普通聊天模态答案得以保留。默认设置按寄存器索引：在一次运行内重新采样（n = 20）发现JSON改变了模型53%的稳定聊天默认设置，大多变回大众答案，并设置了聊天中不存在的默认设置（Claude Fable对颜色0%的时间在聊天中回答“天蓝色”，在JSON中为100%）。全电池控制揭示了一个寄存器梯度：压缩显著且特定于模型训练所使用的答案交付格式（JSON -0.22比特，p =.0002；XML -0.19，p =.002），YAML和CSV不存在，对任意括号包装则相反（+0.13，p =.009）——这将机制倾向于训练后工具使用。在解码器处强制模式（响应格式）的压缩不超过请求（-0.03比特）：这种坍缩存在于模型对寄存器的响应中，而非解码器。结构化输出是软件使用语言模型的方式，与用于评估、比较和选择模型的聊天表面相比，该表面由一个明显更同质的模型提供服务。
英文摘要
When a language model must choose one answer from a large space of equally valid options, a format clause -- "Reply with JSON only" -- changes which answer it chooses. We re-run the One-Word Census (arXiv:2607.12796): 31 wide-answer-space category prompts asked of 44 models, now with the reply requested in JSON -- no schema enforcement, no constrained decoding, only the request. Convergence deepens sharply: on the unconstrained "Pick a word" prompt the modal answer rises from 41% to 64% of the pool and distinct answers fall from 52 to 36; mean answer-choice surprisal drops from 1.80 to 1.58 bits. The tax is progressive: six of 44 models move individually (BH-FDR q=.10), all toward the mode, led by the most distinctive models, while the conformist floor is immobile. It is a sharpener, not a re-indexer -- the plain-chat modal answer survives in 28 of 31 categories. Defaults are register-indexed: a within-run re-sample (n=20) finds JSON shifts 53% of a model's stable chat defaults, mostly back to the crowd, and installs defaults absent from chat (Claude Fable 5 answers "cerulean" for colour 0% of the time in chat, 100% in JSON). Full-battery controls reveal a register gradient: compression is significant and specific to the answer-delivery formats models are trained to speak (JSON -0.22 bits, p=.0002; XML -0.19, p=.002), absent for YAML and CSV, and reversed for an arbitrary bracket wrapper (+0.13, p=.009) -- weighing the mechanism toward tool-use post-training. Enforcing the schema at the decoder (response_format) compresses no further than the request (-0.03 bits): the collapse lives in the model's response to the register, not the decoder. Structured output is how software consumes language models, and that surface is served by a measurably more homogeneous model than the chat surface on which models are evaluated, compared, and chosen.
Comments12 pages, 1 figure. Companion to the One-Word Census (arXiv:2607.12796). Code, data, and interactive explorer: https://github.com/tap2k/modelun/tree/main/studies/structured
