Runtime Fisher Spectral Sensitivity for Early Hallucination Detection

adamzwasserman1 pts0 comments

Runtime Fisher Spectral Sensitivity for Early Hallucination Detection in Edge-Deployed Language Models | Zenodo

Skip to main

You are using an outdated browser. Please upgrade your browser to improve your experience.

Kerimov-Alekberli model

Published July 2, 2026

| Version 1.4

Preprint

Open

Runtime Fisher Spectral Sensitivity for Early Hallucination Detection in Edge-Deployed Language Models

Authors/Creators

Alekberli, Rahid<br>(Researcher)1

Karimov, Hikmat<br>(Researcher)1

Show affiliations

1.

Azerbaijan Technical University

Description

We study whether the spectral sensitivity of the per-token empirical Fisher Information Matrix (FIM), &omega;max(Ft), can serve as a lightweight, model-agnostic runtime signal for anticipating hallucination during autoregressive decoding on consumer edge hardware. Across four pre-registered experiments on twelve open-weight models run locally via Ollama and MLX on an Apple M5 (32GB unified memory), we establish, in turn: cross-run stability of the signal, its predictive correlation with hallucination onset, its early-warning lead time, and the e!ectiveness of an intervention triggered by the resulting alarm. Concretely, ten of twelve models satisfy a stability gate (coe&rdquo;cient of variation Files

KA_FisherEarlyWarning_arxiv_preview.pdf

Files<br>(538.8 kB)

Name<br>Size

Download all

KA_FisherEarlyWarning_arxiv_preview.pdf

md5:96fb318b76a6e46b30a344d79addcc4c

538.8 kB

Preview

Download

Additional details

Software

Repository URL

https://doi.org/10.5281/zenodo.21141999

Programming language

Python

Development Status

Active

References

J. Pennington and P. Worah, "The spectrum of the Fisher information matrix of a single- hidden-layer neural network," in Advances in Neural Information Processing Systems (NeurIPS), 2018. https://proceedings.neurips.cc/paper_files/paper/2018/file/18bb68e2b38e4a8ce7cf4f6b2625768c-Paper.pdf

S. Kadavath et al., "Language models (mostly) know what they know," arXiv preprint arXiv:2207.05221, 2022. https://doi.org/10.48550/arXiv.2207.05221

S. Farquhar et al., "Detecting hallucinations in large language models using semantic entropy," Nature, vol. 630, 2024. https://doi.org/10.1038/s41586-024-07421-0

O. Shorinwa et al., "A survey on uncertainty quantification of large language models: taxonomy, open research challenges, and future directions," arXiv preprint, 2025. https://doi.org/10.48550/arXiv.2412.05563

Views

Downloads

Show more details

All versions<br>This version

Views

Total views

Downloads

Total downloads

Data volume

Total data volume

0 Bytes<br>0 Bytes

More info on how stats are collected....

Versions

External resources

Indexed in

OpenAIRE

Communities

Keywords and subjects

Keywords

Fisher information

hallucination detection

runtime monitoring

on-device inference

quantization

large language models

Details

DOI

DOI Badge

DOI

10.5281/zenodo.21133067

Markdown

[![DOI](https://zenodo.org/badge/DOI/10.5281/zenodo.21133067.svg)](https://doi.org/10.5281/zenodo.21133067)

reStructuredText

.. image:: https://zenodo.org/badge/DOI/10.5281/zenodo.21133067.svg<br>:target: https://doi.org/10.5281/zenodo.21133067

HTML

Image URL

https://zenodo.org/badge/DOI/10.5281/zenodo.21133067.svg

Target URL

https://doi.org/10.5281/zenodo.21133067

Resource type<br>Preprint

Publisher<br>Zenodo

Languages

English

Rights

License

Creative Commons Attribution 4.0 International

The Creative Commons Attribution license allows re-distribution and re-use of a licensed work on the condition that the creator is appropriately credited.

Read more

Copyright

Hikmat Kerimov and Rahid Zahid Alekberli

Citation

Export

Technical metadata

Created

July 2, 2026

Modified

July 2, 2026

Jump up

This site uses cookies. Find out more on how we use cookies

Accept all cookies<br>Accept only essential cookies

zenodo https models language fisher hallucination

Related Articles