`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。
正文是一段关于循环语言模型的文本。我需要仔细阅读并理解内容。
原文:
"Looped language models can stop iterating early, and recent work shows that post-hoc readouts of the loop trajectory often rival the learned exit gate of the looped model Ouro. Such comparisons reduce each method to one operating point chosen by a selection rule. We ask whether that rule decides the outcome. On a frozen Ouro-1.4B with four loops, we compared a trained stability probe ($π_1$-x) with Ouro's native gate, adapted to a per-question stopping decision, in pre-registered studies; the confirmatory study was designed after an exploratory finding on development data and evaluated once on a sealed test set of 8,517 questions. In the pre-registered confirmatory comparison, which selected operating points with a bootstrap-screened 1 pp accuracy-loss rule, the probe saved more of the four-loop budget than the native gate (+0.053 gross macro saving over three benchmarks, 95% CI [+0.049, +0.057]), mostly from CommonsenseQA, where the probe's test upper bound on accuracy loss exceeded 1 pp. Under a point-only rule, a pre-specified secondary analysis on the same test set reversed the preference (-0.022 [-0.026, -0.018]), with operating points that exceeded the budget. In pre-specified secondary comparisons, the probe showed no demonstrated advantage over an untrained KL divergence between successive loops (+0.00125 [-0.0041, +0.0065]); we emphasise this result post hoc and claim no equivalence. The probe also broke its budget on one of two out-of-distribution benchmarks (3.26 pp loss on Social IQa). Savings are estimated from stored full-depth traces, not measured inference time, and our screen carries no formal risk-control guarantee. We conclude that halting methods should be compared as signals together with selection procedures, with strong untrained baselines and out-of-distribution checks."
现在,分析内容:
- 这是一个科学/技术新闻或研究摘要。
- 我需要将它分成自然段落。原文看起来是一个连续的段落,但根据内容,可能可以分成几个部分。
- 规则说:每个自然段落用``包裹。原文似乎是一整段,但为了可读性,我应该考虑是否有自然段落的划分。
仔细看原文:
- 开头:介绍循环语言模型和比较。
- 然后:描述实验设置。
- 接着:描述主要结果。
- 然后:次要分析和结果。
- 最后:结论。
所以,我可以将它分成几个段落。
任务说"每个自然段落",意思是如果原文有多个段落,就用`
`包裹每个。但这里原文是一整段,所以我可能需要基于内容逻辑来划分。
规则2说:"每个自然段落用`
`包裹",所以如果原文是一段,我就用一个`
`。但为了更好的语义,我应该考虑是否有小节。
规则还说:"长文用`
`/`
`划分小节"。这里内容不长,但可能有必要。
原文没有明显的标题,所以可能需要添加小节标题来组织内容。
规则4:保留原文所有信息,不得增删改内容。所以不能添加新内容,但可以添加结构性标签如`
`来划分小节。
规则说"长文用`
`/`
`划分小节",所以对于长文本,我应该添加小节标题。
这个文本长度中等,但既然是新闻正文,可能不需要标题。规则指定了,所以最好添加。
我需要决定如何划分小节。
可能的划分:
1. 介绍部分:关于循环语言模型和比较方法。
2. 实验设置:描述研究设计。
3. 主要结果:确认性比较的结果。
4. 次要分析和结果:点规则下的结果、与未训练基线的比较等。
5. 结论:最终结论。
添加小节标题时,不能增删内容,所以标题应该基于原文内容。
例如:
- 第一段:介绍问题。
- 第二段:实验方法。
- 第三段:主要发现。
- 第四段:次要发现。
- 第五段:结论。
但标题本身是添加的,规则允许用`
`/`
`,所以应该可以。
规则说"不得增删改内容",但添加结构性标签是排版的一部分,不算是增删内容。内容本身不变。
所以,我可以添加小节标题。
现在,检查是否有列表、引用、代码等。
- 没有明显的列表:没有项目符号或编号列表。
- 没有引用:没有直接引用别人的话。
- 有代码吗?原文中有"$π_1$-x",这看起来像LaTeX或数学符号,可能应该用代码标签。
规则3:代码用``。
在文本中,"$π_1$-x",这可能表示数学公式,应该用代码或特殊处理。
在HTML中,数学符号通常用``或特殊标签,但规则指定代码用``。
所以,对于"$π_1$-x",我可以包裹在``中。
同样,CI区间如"[+0.049, +0.057]",可能不需要特殊处理,但可以保持原样。
重要句子:规则说"重要句子可用``",所以我需要识别哪些句子重要。
例如,关键结论或主要发现。
现在,规划HTML结构。
首先,确定小节。
基于内容:
- 开头:介绍循环语言模型的停止问题。
- 然后:实验设置。
- 主要结果。
- 次要结果。
- 结论。
添加小节标题。
例如:
Introduction
但规则说不能增删内容,所以标题应该简洁,基于内容。
或许:
Background
Methodology
Results
Conclusion
但需要确保标题是语义化的。
由于是新闻正文,可能不需要正式标题,但规则要求用``/`
`划分小节。
所以,我会添加。
现在,划分段落。
原文可以分成:
1. "Looped language models can stop iterating early, and recent work shows that post-hoc readouts of the loop trajectory often rival the learned exit gate of the looped model Ouro. Such comparisons reduce each method to one operating point chosen by a selection rule. We ask whether that rule decides the outcome."
2. "On a frozen Ouro-1.4B with four loops, we compared a trained stability probe ($π_1$-x) with Ouro's native gate, adapted to a per-question stopping decision, in pre-registered studies; the confirmatory study was designed after an exploratory finding on development data and evaluated once on a sealed test set of 8,517 questions."
3. "In the pre-registered confirmatory comparison, which selected operating points with a bootstrap-screened 1 pp accuracy-loss rule, the probe saved more of the four-loop budget than the native gate (+0.053 gross macro saving over three benchmarks, 95% CI [+0.049, +0.057]), mostly from CommonsenseQA, where the probe's test upper bound on accuracy loss exceeded 1 pp."
4. "Under a point-only rule, a pre-specified secondary analysis on the same test set reversed the preference (-0.0
`,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。 正文是一段关于循环语言模型的文本。我需要仔细阅读并理解内容。 原文: "Looped language models can stop iterating early, and recent work shows that post-hoc readouts of the loop trajectory often rival the learned exit gate of the looped model Ouro. Such comparisons reduce each method to one operating point chosen by a selection rule. We ask whether that rule decides the outcome. On a frozen Ouro-1.4B with four loops, we compared a trained stability probe ($π_1$-x) with Ouro's native gate, adapted to a per-question stopping decision, in pre-registered studies; the confirmatory study was designed after an exploratory finding on development data and evaluated once on a sealed test set of 8,517 questions. In the pre-registered confirmatory comparison, which selected operating points with a bootstrap-screened 1 pp accuracy-loss rule, the probe saved more of the four-loop budget than the native gate (+0.053 gross macro saving over three benchmarks, 95% CI [+0.049, +0.057]), mostly from CommonsenseQA, where the probe's test upper bound on accuracy loss exceeded 1 pp. Under a point-only rule, a pre-specified secondary analysis on the same test set reversed the preference (-0.022 [-0.026, -0.018]), with operating points that exceeded the budget. In pre-specified secondary comparisons, the probe showed no demonstrated advantage over an untrained KL divergence between successive loops (+0.00125 [-0.0041, +0.0065]); we emphasise this result post hoc and claim no equivalence. The probe also broke its budget on one of two out-of-distribution benchmarks (3.26 pp loss on Social IQa). Savings are estimated from stored full-depth traces, not measured inference time, and our screen carries no formal risk-control guarantee. We conclude that halting methods should be compared as signals together with selection procedures, with strong untrained baselines and out-of-distribution checks." 现在,分析内容: - 这是一个科学/技术新闻或研究摘要。 - 我需要将它分成自然段落。原文看起来是一个连续的段落,但根据内容,可能可以分成几个部分。 - 规则说:每个自然段落用``包裹。原文似乎是一整段,但为了可读性,我应该考虑是否有自然段落的划分。 仔细看原文: - 开头:介绍循环语言模型和比较。 - 然后:描述实验设置。 - 接着:描述主要结果。 - 然后:次要分析和结果。 - 最后:结论。 所以,我可以将它分成几个段落。 任务说"每个自然段落",意思是如果原文有多个段落,就用`
`包裹每个。但这里原文是一整段,所以我可能需要基于内容逻辑来划分。 规则2说:"每个自然段落用`
`包裹",所以如果原文是一段,我就用一个`
`。但为了更好的语义,我应该考虑是否有小节。 规则还说:"长文用`
`/`
`划分小节"。这里内容不长,但可能有必要。 原文没有明显的标题,所以可能需要添加小节标题来组织内容。 规则4:保留原文所有信息,不得增删改内容。所以不能添加新内容,但可以添加结构性标签如`
`来划分小节。 规则说"长文用`
`/`
`划分小节",所以对于长文本,我应该添加小节标题。 这个文本长度中等,但既然是新闻正文,可能不需要标题。规则指定了,所以最好添加。 我需要决定如何划分小节。 可能的划分: 1. 介绍部分:关于循环语言模型和比较方法。 2. 实验设置:描述研究设计。 3. 主要结果:确认性比较的结果。 4. 次要分析和结果:点规则下的结果、与未训练基线的比较等。 5. 结论:最终结论。 添加小节标题时,不能增删内容,所以标题应该基于原文内容。 例如: - 第一段:介绍问题。 - 第二段:实验方法。 - 第三段:主要发现。 - 第四段:次要发现。 - 第五段:结论。 但标题本身是添加的,规则允许用`
`/`
`,所以应该可以。 规则说"不得增删改内容",但添加结构性标签是排版的一部分,不算是增删内容。内容本身不变。 所以,我可以添加小节标题。 现在,检查是否有列表、引用、代码等。 - 没有明显的列表:没有项目符号或编号列表。 - 没有引用:没有直接引用别人的话。 - 有代码吗?原文中有"$π_1$-x",这看起来像LaTeX或数学符号,可能应该用代码标签。 规则3:代码用`
`。 在文本中,"$π_1$-x",这可能表示数学公式,应该用代码或特殊处理。 在HTML中,数学符号通常用``或特殊标签,但规则指定代码用``。 所以,对于"$π_1$-x",我可以包裹在``中。 同样,CI区间如"[+0.049, +0.057]",可能不需要特殊处理,但可以保持原样。 重要句子:规则说"重要句子可用``",所以我需要识别哪些句子重要。 例如,关键结论或主要发现。 现在,规划HTML结构。 首先,确定小节。 基于内容: - 开头:介绍循环语言模型的停止问题。 - 然后:实验设置。 - 主要结果。 - 次要结果。 - 结论。 添加小节标题。 例如:Introduction
但规则说不能增删内容,所以标题应该简洁,基于内容。 或许:Background
Methodology
Results
Conclusion
但需要确保标题是语义化的。 由于是新闻正文,可能不需要正式标题,但规则要求用``/`
`划分小节。 所以,我会添加。 现在,划分段落。 原文可以分成: 1. "Looped language models can stop iterating early, and recent work shows that post-hoc readouts of the loop trajectory often rival the learned exit gate of the looped model Ouro. Such comparisons reduce each method to one operating point chosen by a selection rule. We ask whether that rule decides the outcome." 2. "On a frozen Ouro-1.4B with four loops, we compared a trained stability probe ($π_1$-x) with Ouro's native gate, adapted to a per-question stopping decision, in pre-registered studies; the confirmatory study was designed after an exploratory finding on development data and evaluated once on a sealed test set of 8,517 questions." 3. "In the pre-registered confirmatory comparison, which selected operating points with a bootstrap-screened 1 pp accuracy-loss rule, the probe saved more of the four-loop budget than the native gate (+0.053 gross macro saving over three benchmarks, 95% CI [+0.049, +0.057]), mostly from CommonsenseQA, where the probe's test upper bound on accuracy loss exceeded 1 pp." 4. "Under a point-only rule, a pre-specified secondary analysis on the same test set reversed the preference (-0.0