Catching Contamination Before Generation: Spectral Kill Switches for Agents

TL;DR

This paper presents a spectral analysis method that detects contamination in agentic language models during execution, enabling real-time safety checks without additional training.

Contribution

It introduces a novel spectral statistic-based diagnostic that detects context inconsistency in language models during inference, with theoretical guarantees and practical efficiency.

Findings

High frequency energy ratio effectively detects context inconsistency.

Method operates with less than one millisecond overhead.

Robust bimodality observed across multiple model families.

Abstract

Agentic language models compose multi step reasoning chains, yet intermediate steps can be corrupted by inconsistent context, retrieval errors, or adversarial inputs, which makes post hoc evaluation too late because errors propagate before detection. We introduce a diagnostic that requires no additional training and uses only the forward pass to emit a binary accept or reject signal during agent execution. The method analyzes token graphs induced by attention and computes two spectral statistics in early layers, namely the high frequency energy ratio and spectral entropy. We formalize these signals, establish invariances, and provide finite sample estimators with uncertainty quantification. Under a two regime mixture assumption with a monotone likelihood ratio property, we show that a single threshold on the high frequency energy ratio is optimal in the Bayes sense for detecting context inconsistency. Empirically, the high frequency energy ratio exhibits robust bimodality during context verification across multiple model families, which enables gating decisions with overhead below one millisecond on our hardware and configurations. We demonstrate integration into retrieval augmented agent pipelines and discuss deployment as an inline safety monitor. The approach detects contamination while the model is still processing the text, before errors commit to the reasoning chain.