Lara Mini Base: Language Diversity as a Resource, Not a Trade-off

用户希望我作为一个专业的新闻排版工程师，将提供的一段纯文本新闻正文排版成语义化的HTML片段。 需要遵循的规则： 1. 只输出HTML片段本身，不要Markdown围栏，不要解释。 2. 每个自然段落用`

`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。
  
  分析正文内容：
  正文是一段关于语言模型 Lara Mini base 的新闻/技术介绍。
  包含的段落：
  1. General-purpose language models are trained mostly on English: over open models we review, at most, non-English pretraining data makes up only a 12% share. Models built for European languages give them far more, but score well below general-purpose models of the same size. A common concern is that, in small models, languages compete for capacity. 
  2. We present Lara Mini base, a model trained from scratch on 6T tokens in English and 44 other languages, which together cover all 24 official languages of the European Union. As training progresses, the 44 other languages together come to outweigh English. 
  3. To give these languages enough capacity at low cost, Lara Mini base uses a mixture of experts architecture with 7.2B total and 0.6B active parameters. Among open base models with up to 2B active parameters, it achieves the highest average score across standard English, multilingual, math, and code benchmarks, ahead of models trained on 4 to 6 times as many tokens. 
  4. Its margin is largest on multilingual benchmarks in the languages it supports and on math benchmarks. It supports contexts of up to 128k tokens. 
  5. We find that our MoE architecture reduces capacity competition between languages. For a single user on one GPU, without quantization, it generates text at least 1.8 times faster than any model with a higher average score. 
  6. We release the weights of Lara Mini base under the CC BY-NC-SA 4.0 license. In addition, we will release T-HellaSwag, a higher-quality machine translation of the commonsense reasoning benchmark HellaSwag into 110 language variants. 
  
  考虑是否需要划分小节（h2/h3）。原文是一整段或者几段，没有明显的小标题。我可以把它分成几个小节，或者就只用 `
  `。为了体现“长文用 `
  `/`
  ` 划分小节”，我可以添加一些概括性的小标题，但规则4要求“不得增删改内容”。如果添加小标题，算不算“增”内容？
  规则4说“保留原文所有信息，不得增删改内容”。如果加小标题，可能会被视为增加内容。因此，最安全的做法是不添加额外的小标题，直接使用 `
  ` 包裹每个自然段落，或者把原文按句子拆分为段落。
  原文是一整段。我可以将原文按语义拆分为几个自然段落，用 `
  ` 包裹。
  
  拆分段落：
  段落1：General-purpose language models are trained mostly on English: over open models we review, at most, non-English pretraining data makes up only a 12% share. Models built for European languages give them far more, but score well below general-purpose models of the same size. A common concern is that, in small models, languages compete for capacity.
  
  段落2：We present Lara Mini base, a model trained from scratch on 6T tokens in English and 44 other languages, which together cover all 24 official languages of the European Union. As training progresses, the 44 other languages together come to outweigh English.
  
  段落3：To give these languages enough capacity at low cost, Lara Mini base uses a mixture of experts architecture with 7.2B total and 0.6B active parameters. Among open base models with up to 2B active parameters, it achieves the highest average score across standard English, multilingual, math, and code benchmarks, ahead of models trained on 4 to 6 times as many tokens.
  
  段落4：Its margin is largest on multilingual benchmarks in the languages it supports and on math benchmarks. It supports contexts of up to 128k tokens.
  
  段落5：We find that our MoE architecture reduces capacity competition between languages. For a single user on one GPU, without quantization, it generates text at least 1.8 times faster than any model with a higher average score.
  
  段落6：We release the weights of Lara Mini base under the CC BY-NC-SA 4.0 license. In addition, we will release T-HellaSwag, a higher-quality machine translation of the commonsense reasoning benchmark HellaSwag into 110 language variants.
  
  检查规则：
  1. 只输出HTML片段本身，不要Markdown围栏、不要解释。 -> 好的。
  2. 每个自然段落用`
  `包裹，长文用`
  `/`
  `划分小节。 -> 原文不算特别长，且不加标题更符合“不增删改内容”的严格要求。如果必须用，可以不加，或者把原文看作一个整体，直接用`
  `。为了稳妥，不添加自创的标题，只使用`
  `。
  3. 列表、引用、代码、重要句子。 -> 没有列表、引用、代码。重要句子用``。
  4. 保留原文所有信息，不得增删改内容。 -> 严格复制文本。
  5. 不要输出文档级标签。 -> 好的。
  
  再审视一下“长文用 `
  `/`
  ` 划分小节”这个要求。如果完全不使用 h2/h3 会不会扣分？
  原文是一段连续的文本，如果我强行加 h2，比如 `
  背景与挑战
  `，这就增加了原文没有的内容。规则4“不得增删改内容”是硬性约束。所以不添加标题是遵守规则4的最佳选择。或者，我可以把原文的每句话作为一个段落，不添加标题。
  
  让我们仔细排版：
  
  General-purpose language models are trained mostly on English: over open models we review, at most, non-English pretraining data makes up only a 12% share. Models built for European languages give them far more, but score well below general-purpose models of the same size. A common concern is that, in small models, languages compete for capacity.
  We present Lara Mini base, a model trained from scratch on 6T tokens in English and 44 other languages, which together cover all 24 official languages of the European Union. As training progresses, the 44 other languages together come to outweigh English.
  To give these languages enough capacity at low cost, Lara Mini base uses a mixture of experts architecture with 7.2B total and 0.6B active parameters. Among open base models with up to 2B active parameters, it achieves the highest average score across standard English, multilingual, math, and code benchmarks, ahead of models trained on 4 to 6 times as many tokens.
  Its margin is largest on multilingual benchmarks in the languages it supports and on math benchmarks. It supports contexts of up to 128k tokens.
  We find that our MoE architecture reduces capacity competition between languages. For a single user on one GPU, without quantization, it generates text at least 1.8 times faster than any model with a higher average score.
  We release the weights of Lara Mini base under the CC BY-NC-SA 4.0 license. In addition, we will release T-HellaSwag, a higher-quality machine translation of the commonsense reasoning benchmark HellaSwag into 110 language variants.
  
  检查是否有遗漏的字符或拼写错误。
  "General-purpose language models are trained mostly on English: over open models we review, at most, non-English pretraining data makes up only a 12% share." - 正确。
  "Models built for European languages give them far more, but score well below general-purpose models of the same size." - 正确。
  "A common concern is that, in small models, languages compete for