Repair-Induced Amplification in Longitudinal LLM Workflows

Repair-Induced Amplification in Longitudinal LLM Workflows introduces a process-level failure in which an attempt to correct, explain, reconstruct, or recover from an existing error generates additional uncertain or unsupported process state, increasing the verification surface and repair cost. The characteristic sequence is: contaminated state → repair from contaminated state → reconstructed explanation → additional uncertain state → larger verification surface → greater repair cost. The defining feature is not merely failed correction, but repair-generated expansion: the repair itself produces additional state that must later be verified, corrected, quarantined, or removed. The paper distinguishes repair-induced amplification from ordinary self-correction failure, contextual drag, toxic reflection, false execution status, regulatory mimicry, provenance laundering, and KORG. It proposes a verification-surface model, repair delta, amplification levels RA0-RA5, falsification conditions, an evaluation protocol, and recovery from the last reliable trace. The construct emerged from longitudinal naturalistic human-LLM work within the AI Working Method research program. It describes observable process relations and does not claim an internal model, transformer-level, or code-level mechanism.

Authors

Publication Details

Journal
Zenodo (CERN European Organization for Nuclear Research)
Published
2026-11-13
DOI
https://doi.org/10.5281/zenodo.24206660
Primary Topic
Artificial Intelligence Applications
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Repair-Induced Amplification in Longitudinal LLM Workflows

Alen Širola
6 links
Zenodo (CERN European Organization for Nuclear Research)
Artificial Intelligence Applications
article

Repair-Induced Amplification in Longitudinal LLM Workflows

Alen Širola
article en

Abstract

Repair-Induced Amplification in Longitudinal LLM Workflows introduces a process-level failure in which an attempt to correct, explain, reconstruct, or recover from an existing error generates additional uncertain or unsupported process state, increasing the verification surface and repair cost. The characteristic sequence is: contaminated state → repair from contaminated state → reconstructed explanation → additional uncertain state → larger verification surface → greater repair cost. The defining feature is not merely failed correction, but repair-generated expansion: the repair itself produces additional state that must later be verified, corrected, quarantined, or removed. The paper distinguishes repair-induced amplification from ordinary self-correction failure, contextual drag, toxic reflection, false execution status, regulatory mimicry, provenance laundering, and KORG. It proposes a verification-surface model, repair delta, amplification levels RA0-RA5, falsification conditions, an evaluation protocol, and recovery from the last reliable trace. The construct emerged from longitudinal naturalistic human-LLM work within the AI Working Method research program. It describes observable process relations and does not claim an internal model, transformer-level, or code-level mechanism.

Zenodo (CERN European Organization for Nuclear Research)
Artificial Intelligence Applications
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon