首页 > AI前沿 > Can an LLM dream? Testing the overfitted brain hypothesis on an agent with text memory

Can an LLM dream? Testing the overfitted brain hypothesis on an agent with text memory

arXiv自然语言 2026-11-18 13:00 18 阅读 查看原文
用户要求将一段纯文本新闻正文排版成语义化的HTML片段。 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。
    
    正文内容:
    Hoel (2021) proposed that dreams distort replayed experience so that the brain does not overfit to recent events. We test this hypothesis on an LLM agent (Claude Opus 5.5) that stores what it learns in a text memory and rewrites that memory each night from the day's conversation transcript. The agent goes through ten simulated days of conversation built from LongMemEval. We corrupt 5, 10, 20 or 40 percent of the transcript's turns before each nightly rewrite in one of two ways. Detail corruption changes one or two details in a turn to plausible wrong ones, or replaces the turn with one by the same speaker from an earlier day. Random-text corruption replaces half of the selected turns with a passage of Moby-Dick and inserts Moby-Dick words into the other half. A size limit on the memory, rising from 19,500 to 117,000 tokens over ten nights, set from human free-recall data, forces it to forget. We learned that corruption did not help the agent apply what it learned in earlier conversations to new tasks in any of 30 runs (ten conditions, three different conversation histories). Replacing details in 40 percent of turns lowered recall from 76 to 60 percent. Random text had no consistent effect on recall, with a paired difference of 0 $\pm$ 2 points at 40 percent, because the model recognized it as corrupted and left it out of memory. Replaced details were stored as fact even when another turn held the correct value, and the agent answered 40 percent of those questions correctly, compared with 75 percent without corruption. These results suggest that a pretrained model can separate corrupted text from real events, so a fair test needs a learner that cannot make this separation as easily, such as a model fine-tuned on the corrupted days or an agent whose days take place in an invented world.
    
    分析正文结构,这是一段连续的学术新闻/摘要,没有明显的列表或引用。可以将其划分为几个逻辑小节,或者就当成一个长段落,但规则说“长文用`

    `/`

    `划分小节”。 可以划分成: `

    ` 研究背景与假设 `

    ` Hoel (2021) proposed... `

    ` 实验方法 `

    ` We test this hypothesis... `

    ` The agent goes through ten simulated days... `

    ` We corrupt 5, 10, 20 or 40 percent... `

    ` Detail corruption changes... `

    ` Random-text corruption replaces... `

    ` A size limit on the memory... `

    ` 实验结果 `

    ` We learned that corruption did not help... `

    ` Replacing details in 40 percent... `

    ` Random text had no consistent effect... `

    ` Replaced details were stored as fact... `

    ` 结论 `

    ` These results suggest that... 等等,规则要求“保留原文所有信息,不得增删改内容”。如果我加了`

    `小节标题,是不是算“增删改内容”? 规则说:“保留原文所有信息,不得增删改内容”,通常指文本内容不能改。加HTML标签是排版,但如果加了文本标题(如“研究背景”),那就是增加了文本内容。所以,为了严格遵守“不得增删改内容”,我不能添加原文没有的标题文本。 那怎么用`

    `/`

    `划分小节呢?也许不需要强制使用,或者只用在原文本身有标题的情况下(但原文没有)。 如果不用`

    `/`

    `,直接全部用`

    `包裹,也是符合“每个自然段落用`

    `包裹”的。原文其实可以视为几个自然段落,或者就是一个大段落。 为了安全,直接把原文按句子或逻辑分成几个`

    `,不添加任何额外的文本标签内容,不添加`

    `以免引入新文字。或者,我可以把第一句话作为`

    `?不行,那会改变文本的呈现方式且可能被视为修改。 最稳妥的做法:把原文分成几个自然段,用`

    `包裹。如果非要用`

    `,只能把原文的句子直接放进`

    `,但这不符合常规。 其实原文就是一段,或者我按句群分成几个`

    `。 原文: Hoel (2021) proposed that dreams distort replayed experience so that the brain does not overfit to recent events. We test this hypothesis on an LLM agent (Claude Opus 5.5) that stores what it learns in a text memory and rewrites that memory each night from the day's conversation transcript. The agent goes through ten simulated days of conversation built from LongMemEval. We corrupt 5, 10, 20 or 40 percent of the transcript's turns before each nightly rewrite in one of two ways. Detail corruption changes one or two details in a turn to plausible wrong ones, or replaces the turn with one by the same speaker from an earlier day. Random-text corruption replaces half of the selected turns with a passage of Moby-Dick and inserts Moby-Dick words into the other half. A size limit on the memory, rising from 19,500 to 117,000 tokens over ten nights, set from human free-recall data, forces it to forget. We learned that corruption did not help the agent apply what it learned in earlier conversations to new tasks in any of 30 runs (ten conditions, three different conversation histories). Replacing details in 40 percent of turns lowered recall from 76 to 60 percent. Random text had no consistent effect on recall, with a paired difference of 0 $\pm$ 2 points at 40 percent, because the model recognized it as corrupted and left it out of memory. Replaced details were stored as fact even when another turn held the correct value, and the agent answered 40 percent of those questions correctly, compared with 75 percent without corruption. These results suggest that a pretrained model can separate corrupted text from real events, so a fair test needs a learner that cannot make this separation as easily, such as a model fine-tuned on the corrupted days or an agent whose days take place in an invented world. 这样分成5个`

    `。不需要`

      `、`
      `、`
      `,因为原文没有这些内容。重要句子可以用``,比如结论部分,或者不加,规则说“可用”,不是“必须”。
      为了体现“重要句子可用``”,我可以把最后一句或者核心结论加``。
      比如:These results suggest that a pretrained model can separate corrupted text from real events, so a fair test needs a learner that cannot make this separation as easily, such as a model fine-tuned on the corrupted days or an agent whose days take place in an invented world.
      
      检查规则:
      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;(没有长文标题,所以不用`

      `/`

      `,或者把某些句子当作标题?不,不添加新文本最安全。如果必须用,可以把第一句当`

      `?不,第一句不是标题。直接用`

      `即可。) 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。
        
        排版结果