CogniConsole: Externalizing Inference-Time Control as a Formal Abstraction for Reliable LLM Interactions

CogniConsole：将推理时控制外化为可靠大语言模型交互的形式化抽象

Abstract: Reliability in large language model (LLM) systems is typically framed as a function of model capability. We challenge this by demonstrating that reliability is significantly influenced by \emph{inference-time control} — the computational layer governing task framing and context selection.

摘要： 大语言模型（LLM）系统的可靠性通常被视为模型能力的函数。我们对此提出了挑战，并证明可靠性在很大程度上受到“推理时控制”（inference-time control）的影响——即管理任务框架和上下文选择的计算层。

We introduce \emph{CogniConsole}, an architectural instantiation that externalizes this control into a structured interface combining programmatic coordination with bounded prompt-based reasoning.

我们引入了“CogniConsole”，这是一种架构实例化方案，它将这种控制外化为一个结构化接口，结合了程序化协调与有界提示词推理。

Through \emph{controllability-oriented probes} ($N=489$) in a multi-step interactive environment, we show that increasing structural scaffolding — from unstructured to fully scaffolded — \textbf{systematically reduces output variance and failure rates under a fixed model architecture}.

通过在多步交互环境中的“可控性导向探测”（$N=489$），我们展示了增加结构化脚手架（从非结构化到完全脚手架化）能够在固定模型架构下系统性地降低输出方差和故障率。

Our results indicate that many observed failure modes, such as context drift and inconsistent constraint adherence, arise from under-specified control rather than insufficient capability.

我们的研究结果表明，许多观察到的故障模式（如上下文漂移和约束遵循不一致）源于控制定义不足，而非模型能力不足。

This work provides an empirical basis for treating inference-time control as a first-class abstraction, opening new directions for designing and evaluating LLM systems beyond scaling alone.

这项工作为将推理时控制视为一等抽象提供了实证基础，为设计和评估大语言模型系统开辟了超越单纯扩展规模之外的新方向。