包裹，长文用
/
划分小节。 3. 列表内容用
- ，引用用
，代码用
，重要句子可用。
4. 保留原文所有信息，不得增删改内容。
5. 不要输出//等文档级标签。
正文是：
Decision models read a document and return a probability for each of a fixed set of options instead of generating text. Several appeared in 2026, but none has published results outside English. We describe Zuse, three decision models (4B, 9B and 27B) that train a low-rank adapter and a pointer head on frozen open-weight bases, targeting 27 European languages, and use them to study how this design behaves across languages and tasks. On a broad European suite of 11 sources excluded from training, covering 9 task groups and 27 languages, a development set that also guided model selection, the 27B model reaches 87.5% accuracy with a pooled expected calibration error of 0.016, and 0.035 when weighted per source. Post-hoc analyses of these single-seed results give three findings. First, miscalibration is structured by task rather than language: on parallel reading and topic items in 26 languages the confidence gap varies about as much as sampling noise, while across sources it spans -0.07 to +0.09, with knowledge tasks overconfident and acceptability judgements underconfident at all three sizes, so one fitted temperature cannot correct both. Second, the share of decisions that can be automated at a 5% error budget follows chance-corrected skill (Spearman 0.97 over 33 source and model pairs) more closely than raw accuracy (0.91). Third, reach beyond the training languages is set by the base model: across 210 languages, changing the training run moved scores by at most about 2 points within a size, while languages outside the base's documented list trail by 23 to 28 points at every size. On English public suites Zuse is behind Jev and Kev, including Kev models on the same bases. We also report two pitfalls in building multi-source training data: label-word collision between topic schemes, and an injection augmentation that taught a shortcut. Zuse can be tried at https://bineric.com/zuse.
我需要分析这个文本，确定如何分割成段落、小节等。
文本结构：
- 开头介绍决策模型。
- 然后描述Zuse模型。
- 接着是实验结果。
- 之后是三个主要发现。
- 最后是关于陷阱和试用链接。
规则说：每个自然段落用包裹。长文用
/
划分小节。
所以，我应该将文本分成逻辑段落。
让我试着分段：
1. 第一句："Decision models read a document and return a probability for each of a fixed set of options instead of generating text." 这是一个自然段落。
2. 第二句："Several appeared in 2026, but none has published results outside English." 可能和上面连在一起，但可能是一个新段落。看上下文。
实际上，整个文本是连续的，但可以分成几个部分。
让我读一下：
- "Decision models read a document and return a probability for each of a fixed set of options instead of generating text." – 介绍。
- "Several appeared in 2026, but none has published results outside English." – 继续。
- "We describe Zuse, three decision models (4B, 9B and 27B) that train a low-rank adapter and a pointer head on frozen open-weight bases, targeting 27 European languages, and use them to study how this design behaves across languages and tasks." – 描述Zuse。
- "On a broad European suite of 11 sources excluded from training, covering 9 task groups and 27 languages, a development set that also guided model selection, the 27B model reaches 87.5% accuracy with a pooled expected calibration error of 0.016, and 0.035 when weighted per source." – 实验设置和结果。
- "Post-hoc analyses of these single-seed results give three findings." – 引出发现。
- "First, miscalibration is structured by task rather than language: on parallel reading and topic items in 26 languages the confidence gap varies about as much as sampling noise, while across sources it spans -0.07 to +0.09, with knowledge tasks overconfident and acceptability judgements underconfident at all three sizes, so one fitted temperature cannot correct both." – 第一个发现。
- "Second, the share of decisions that can be automated at a 5% error budget follows chance-corrected skill (Spearman 0.97 over 33 source and model pairs) more closely than raw accuracy (0.91)." – 第二个发现。
- "Third, reach beyond the training languages is set by the base model: across 210 languages, changing the training run moved scores by at most about 2 points within a size, while languages outside the base's documented list trail by 23 to 28 points at every size." – 第三个发现。
- "On English public suites Zuse is behind Jev and Kev, including Kev models on the same bases." – 其他结果。
- "We also report two pitfalls in building multi-source training data: label-word collision between topic schemes, and an injection augmentation that taught a shortcut." – 陷阱。
- "Zuse can be tried at https://bineric.com/zuse." – 试用链接。
现在，为了HTML排版：
- 每个自然段落用
包裹。所以，我应该将文本分成几个
。
- 长文用
，代码用，重要句子可用。 4. 保留原文所有信息，不得增删改内容。 5. 不要输出//等文档级标签。 正文是： Decision models read a document and return a probability for each of a fixed set of options instead of generating text. Several appeared in 2026, but none has published results outside English. We describe Zuse, three decision models (4B, 9B and 27B) that train a low-rank adapter and a pointer head on frozen open-weight bases, targeting 27 European languages, and use them to study how this design behaves across languages and tasks. On a broad European suite of 11 sources excluded from training, covering 9 task groups and 27 languages, a development set that also guided model selection, the 27B model reaches 87.5% accuracy with a pooled expected calibration error of 0.016, and 0.035 when weighted per source. Post-hoc analyses of these single-seed results give three findings. First, miscalibration is structured by task rather than language: on parallel reading and topic items in 26 languages the confidence gap varies about as much as sampling noise, while across sources it spans -0.07 to +0.09, with knowledge tasks overconfident and acceptability judgements underconfident at all three sizes, so one fitted temperature cannot correct both. Second, the share of decisions that can be automated at a 5% error budget follows chance-corrected skill (Spearman 0.97 over 33 source and model pairs) more closely than raw accuracy (0.91). Third, reach beyond the training languages is set by the base model: across 210 languages, changing the training run moved scores by at most about 2 points within a size, while languages outside the base's documented list trail by 23 to 28 points at every size. On English public suites Zuse is behind Jev and Kev, including Kev models on the same bases. We also report two pitfalls in building multi-source training data: label-word collision between topic schemes, and an injection augmentation that taught a shortcut. Zuse can be tried at https://bineric.com/zuse. 我需要分析这个文本，确定如何分割成段落、小节等。 文本结构： - 开头介绍决策模型。 - 然后描述Zuse模型。 - 接着是实验结果。 - 之后是三个主要发现。 - 最后是关于陷阱和试用链接。 规则说：每个自然段落用包裹。长文用
/
划分小节。 所以，我应该将文本分成逻辑段落。 让我试着分段： 1. 第一句："Decision models read a document and return a probability for each of a fixed set of options instead of generating text." 这是一个自然段落。 2. 第二句："Several appeared in 2026, but none has published results outside English." 可能和上面连在一起，但可能是一个新段落。看上下文。 实际上，整个文本是连续的，但可以分成几个部分。 让我读一下： - "Decision models read a document and return a probability for each of a fixed set of options instead of generating text." – 介绍。 - "Several appeared in 2026, but none has published results outside English." – 继续。 - "We describe Zuse, three decision models (4B, 9B and 27B) that train a low-rank adapter and a pointer head on frozen open-weight bases, targeting 27 European languages, and use them to study how this design behaves across languages and tasks." – 描述Zuse。 - "On a broad European suite of 11 sources excluded from training, covering 9 task groups and 27 languages, a development set that also guided model selection, the 27B model reaches 87.5% accuracy with a pooled expected calibration error of 0.016, and 0.035 when weighted per source." – 实验设置和结果。 - "Post-hoc analyses of these single-seed results give three findings." – 引出发现。 - "First, miscalibration is structured by task rather than language: on parallel reading and topic items in 26 languages the confidence gap varies about as much as sampling noise, while across sources it spans -0.07 to +0.09, with knowledge tasks overconfident and acceptability judgements underconfident at all three sizes, so one fitted temperature cannot correct both." – 第一个发现。 - "Second, the share of decisions that can be automated at a 5% error budget follows chance-corrected skill (Spearman 0.97 over 33 source and model pairs) more closely than raw accuracy (0.91)." – 第二个发现。 - "Third, reach beyond the training languages is set by the base model: across 210 languages, changing the training run moved scores by at most about 2 points within a size, while languages outside the base's documented list trail by 23 to 28 points at every size." – 第三个发现。 - "On English public suites Zuse is behind Jev and Kev, including Kev models on the same bases." – 其他结果。 - "We also report two pitfalls in building multi-source training data: label-word collision between topic schemes, and an injection augmentation that taught a shortcut." – 陷阱。 - "Zuse can be tried at https://bineric.com/zuse." – 试用链接。 现在，为了HTML排版： - 每个自然段落用
包裹。所以，我应该将文本分成几个
。 - 长文用