事实缺失的地方：气隙式大语言模型智能体管道中信息遗漏的分层分类和逐层归因
Where Facts Go Missing: A Layerwise Taxonomy and Per-Layer Attribution of Information Omission in Air-Gapped LLMAgent Pipelines
浏览论文内容
中文总结 AI 辅助
研究气隙式大语言模型智能体管道中信息遗漏问题，提出九层分类法、归因方法、跨架构比较框架及运行时检测框架，经多模型多引擎试验，确定遗漏率及来源，为定位运营商干预位置提供依据。
中文摘要 AI 辅助
在受监管环境（临床FHIR服务、法律审查、主权基础设施）中的气隙式和本地部署无法调用前沿API；它们通过此http URL或工具服务器后面的vLLM运行量化的4 - 8B模型。主要的可靠性故障是遗漏：关键决策事实的无声缺失，例如智能体读取400条记录中的20条并报告“无异常”。我们认为遗漏是管道现象而非模型现象，并做出四项贡献。一是九层分类法（L0 - L8）定位从摄取到智能体循环的每个遗漏机制；二是归因方法，通过可控消融和逻辑分解将确定性层（L0 - L3）与行为层（L4 - L8）分离，并用遗漏瀑布图量化；三是跨架构比较框架，比较不同引擎和框架的滑动窗口混合、全注意力和SSM混合模型；四是气隙设置的运行时检测框架。对五个模型和两个引擎进行75476次试验的结果显示合并遗漏率为0.62；68%源于确定性中间件（L0 - L3），重新定位了运营商应干预的位置。服务器端配置因素（权重量化、KV缓存类型、RoPE缩放）固定留待未来研究。
英文摘要
Air-gapped and on-premises language-model agents can silently omit decision-critical facts at any boundary between source ingestion and final answer generation. We present a nine-layer taxonomy (L0-L8), an instrumented attribution harness, and a conditional omission waterfall that distinguishes deterministic software loss from behavioral non-retrieval. We analyze 75,476 controlled synthetic trials spanning five open-weight model configurations and two inference engines, together with a separate 372-trial real-agent pilot covering FHIR, PubMed, and SEC-EDGAR sources with LangChain and ADK orchestration. The weighted synthetic benchmark yields an omission rate of 0.574 (95% CI: 0.571-0.578); deliberately injected deterministic faults at L0-L3 account for 73.4% of weighted loss under the benchmark allocation. Increasing context length is most strongly associated with omission (odds ratio 7.43, 95% CI: 5.44-10.15). Completed server-profile analyses associate q4 KV cache and scaled RoPE with higher omission. In the real-agent pilot, 57.8% of traces are unsuccessful overall and 50.9% remain unsuccessful after excluding execution errors. These results establish pipeline-level attribution in a controlled stress test, but benchmark allocations, confounded model comparisons, and heuristic behavioral labels do not measure production prevalence or causal architectural effects.
