Research

Paper

TESTING March 17, 2026

Noise-Response Calibration: A Causal Intervention Protocol for LLM-Judges

Authors

Maxim Khomiakov, Jes Frellsen

Abstract

Large language models (LLMs) are increasingly used as automated judges and synthetic labelers, especially in low-label settings. Yet these systems are stochastic and often overconfident, which makes deployment decisions difficult when external ground truth is limited. We propose a practical calibration protocol based on controlled input interventions: if noise severity increases, task performance should exhibit a statistically significant deterioration trend. We operationalize this with a slope-based hypothesis test over repeated trials, using signal-to-noise-ratio (SNR) perturbations for tabular data and lexical perturbations for text data. Across UCI tabular benchmarks and four text classification datasets, we find clear modality-dependent behavior. Our results reveal a modality gap: while text-based judges degrade predictably, the majority of tabular datasets show a lack of statistically significant performance deterioration even under significant signal-to-noise reduction. Interestingly we find that model performance is lower on datasets that are insensitive to noise interventions. We present a reproducible methodology and reporting protocol for robust LLM-judge calibration under distribution shift.

Metadata

arXiv ID: 2603.17172
Provider: ARXIV
Primary Category: cs.LG
Published: 2026-03-17
Fetched: 2026-03-19 06:01

Related papers

Raw Data (Debug)
{
  "raw_xml": "<entry>\n    <id>http://arxiv.org/abs/2603.17172v1</id>\n    <title>Noise-Response Calibration: A Causal Intervention Protocol for LLM-Judges</title>\n    <updated>2026-03-17T22:08:06Z</updated>\n    <link href='https://arxiv.org/abs/2603.17172v1' rel='alternate' type='text/html'/>\n    <link href='https://arxiv.org/pdf/2603.17172v1' rel='related' title='pdf' type='application/pdf'/>\n    <summary>Large language models (LLMs) are increasingly used as automated judges and synthetic labelers, especially in low-label settings. Yet these systems are stochastic and often overconfident, which makes deployment decisions difficult when external ground truth is limited. We propose a practical calibration protocol based on controlled input interventions: if noise severity increases, task performance should exhibit a statistically significant deterioration trend. We operationalize this with a slope-based hypothesis test over repeated trials, using signal-to-noise-ratio (SNR) perturbations for tabular data and lexical perturbations for text data. Across UCI tabular benchmarks and four text classification datasets, we find clear modality-dependent behavior. Our results reveal a modality gap: while text-based judges degrade predictably, the majority of tabular datasets show a lack of statistically significant performance deterioration even under significant signal-to-noise reduction. Interestingly we find that model performance is lower on datasets that are insensitive to noise interventions. We present a reproducible methodology and reporting protocol for robust LLM-judge calibration under distribution shift.</summary>\n    <category scheme='http://arxiv.org/schemas/atom' term='cs.LG'/>\n    <published>2026-03-17T22:08:06Z</published>\n    <arxiv:comment>Published as a conference paper at CAO Workshop at ICLR 2026</arxiv:comment>\n    <arxiv:primary_category term='cs.LG'/>\n    <author>\n      <name>Maxim Khomiakov</name>\n    </author>\n    <author>\n      <name>Jes Frellsen</name>\n    </author>\n  </entry>"
}