首页 > AI前沿 > LMLineage: Content Lineage and Context Governance for LLM Agents

LMLineage: Content Lineage and Context Governance for LLM Agents

arXiv自然语言 2026-11-17 18:05 14 阅读 查看原文
用户需要将一段纯文本新闻正文排版成语义化的HTML片段。需要仔细检查原文,识别出哪些是标题、段落、列表、引用、代码等。原文是一段关于LMLineage系统的技术描述。 规则要求: 1. 只输出HTML片段,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。
    
    分析原文:
    原文整体是一个连贯的段落,描述了研究背景、提出的系统、方法、实验和结果。没有明显的列表、引用、代码或明显的小节标题。不过可以根据内容逻辑划分小节,比如背景与系统、方法、评估与结果、补充说明。
    
    原文内容拆解:
    "Interpretability for AI agents should not be confined to understanding the model's internal weights. It must also cover the agent's reasoning and decision-making process over multi-source context. We present LMLineage, an out-of-band system that attributes every output sentence of an agent session to its supporting input spans with calibrated confidences, without requiring white-box access to the model or any additional inference passes." -> 这是背景和系统介绍。可以作为第一段,或者加个小标题。
    
    "On top of the lineage metadata, we build concept-drift micro compaction: instead of waiting for the context-length limit and summarizing everything, the system detects when previously used input documents have stopped being referenced ("drifted") and surgically compacts only those, preserving user instructions verbatim." -> 这是方法介绍。可以作为第二段,或者加小标题。
    
    "We evaluate on a multi-tier corpus: 71 gold annotated sentences from 10 Chinese web-QA cases, plus 90 turns of independently collected multi-turn sessions across three generator models (gpt-4o-mini, qwen3.8-27b, deepseek-v4-flash). Results show that (i) window-level embedding similarity is far better aligned with LLM paraphrasing than whole document or lexical matching; (ii) the greedy-window fast path is both faster and more accurate on average than exhaustive line-matching (a multiple-comparisons bias in the latter inflates web recall at the cost of overall accuracy); (iii) per-turn attribution completes in 1s on a single GPU, remaining stable as context accumulates to 40 turns; and (iv) concept-drift detection is conservative (0% false compaction) with recall bounded by attribution accuracy." -> 这是评估和结果。这里有一个结果列表 (i) (ii) (iii) (iv)。可以考虑把结果部分转换为列表,或者保持段落但加粗重要部分。原文用的是 (i) (ii) 这种形式,不是严格的列表,但为了语义化,可以将其转为
      列表,或者保留在段落中。考虑到规则3提到列表内容用
      • ,这里的结果用 (i) (ii) (iii) (iv) 列出,可以将其提取为列表。但要注意"不得增删改内容"。如果提取为列表,需要把 (i) 等符号去掉或保留?通常可以保留文本,去掉 (i) 这种前缀,或者保留。为了严格不增删,可以将其作为列表项,保留原文的 (i) 等,或者直接写成列表项。更好的做法是保留原文的段落结构,把 (i) 到 (iv) 放在一个段落里,或者将其拆分为列表。如果拆分为列表,需要把前面的"Results show that"保留。例如:

        We evaluate on a multi-tier corpus: 71 gold annotated sentences from 10 Chinese web-QA cases, plus 90 turns of independently collected multi-turn sessions across three generator models (gpt-4o-mini, qwen3.8-27b, deepseek-v4-flash). Results show that:

        • (i) window-level embedding similarity is far better aligned with LLM paraphrasing than whole document or lexical matching;
        • (ii) the greedy-window fast path is both faster and more accurate on average than exhaustive line-matching (a multiple-comparisons bias in the latter inflates web recall at the cost of overall accuracy);
        • (iii) per-turn attribution completes in 1s on a single GPU, remaining stable as context accumulates to 40 turns; and
        • (iv) concept-drift detection is conservative (0% false compaction) with recall bounded by attribution accuracy.
        这样既保留了内容,又语义化了。但原文是连在一起的,拆开是否算"增删改"?通常排版允许合理拆分。为了安全,也可以保持段落,但用强调。不过规则鼓励列表内容用
        • ,这里明显是列表形式。 最后一段:"We further describe the concept-drift micro compaction design and an interactive incident-tracing/memory-cleaning case study. A companion patent application covers the lineage construction method." -> 补充说明。 是否需要加小标题?原文没有标题,但规则2说"长文用

          /

          划分小节"。这篇不算很长,但为了结构清晰,可以加几个小标题。例如:

          背景与系统

          ...

          方法:概念漂移微压缩

          ...

          评估与结果

          ...

            ...

          补充说明

          ...

          但是加小标题属于"增"内容,规则4说"不得增删改内容"。如果加小标题,就是增加了原文没有的标题文字。因此,最好不要加

          /

          标题,除非原文本身有标题。原文是一段连续的文本,没有明显的标题。所以应该只用

          包裹,或者将结果部分转为列表。 再仔细看原文,第一句到"inference passes."是一段。第二句"On top of..."到"verbatim."是一段。第三句"We evaluate..."到"accuracy."是一段,其中包含结果。最后一句"We further describe..."到"method."是一段。 所以可以这样排版:

          Interpretability for AI agents should not be confined to understanding the model's internal weights. It must also cover the agent's reasoning and decision-making process over multi-source context. We present LMLineage, an out-of-band system that attributes every output sentence of an agent session to its supporting input spans with calibrated confidences, without requiring white-box access to the model or any additional inference passes.

          On top of the lineage metadata, we build concept-drift micro compaction: instead of waiting for the context-length limit and summarizing everything, the system detects when previously used input documents have stopped being referenced ("drifted") and surgically compacts only those, preserving user instructions verbatim.

          We evaluate on a multi-tier corpus: 71 gold annotated sentences from 10 Chinese web-QA cases, plus 90 turns of independently collected multi-turn sessions across three generator models (gpt-4o-mini, qwen3.8-27b, deepseek-v4-flash). Results show that:

          • (i) window-level embedding similarity is far better aligned with LLM paraphrasing than whole document or lexical matching;
          • (ii) the greedy-window fast path is both faster and more accurate on average than exhaustive line-matching (a multiple-comparisons bias in the latter inflates web recall at the cost of overall accuracy);
          • (iii) per-turn attribution completes in 1s on a single GPU, remaining stable as context accumulates to 40 turns; and
          • (iv) concept-drift detection is conservative (0% false compaction) with recall bounded by attribution accuracy.

          We further describe the concept-drift micro compaction design and an interactive incident-tracing/memory-cleaning case study. A companion patent application covers the lineage construction method.

          关于重要句子用:可以加粗"LMLineage"和"concept-drift micro compaction"等关键术语,但不要