Stopping as a Database Property: Reason-Bearing Termination for Long-Running LLM Agents

Loop termination should be a property of the data layer: when the substrate owns resource and intent state, stopping becomes a reason-bearing, auditable predicate rather than an arbitrary step cap or an unreliable model decision. We document this in a self-built agent research testbed. The substrate's signal ledger holds 48,061 self-monitoring events over 45 days; loop-related entries reference round counts up to 212 (208 distinct values). A mutual-critique relay loop emerged on its own and converged naturally at 29 rounds, seven weeks before any predicate formalized that behavior. A self-narration layer grew to 12,102 entries. On the official ARC-AGI-3 benchmark: 28.57 (3/6 levels, 31 steps) and 0.15 on Kaggle, operating under a self-imposed 82-step budget. We crystallize the mechanisms into four data-computable termination predicates, and document a recurring dissociation: convergence control and converged-to content are separable—a relay loop faithfully amplifies upstream wiring faults. Contributions: placement taxonomy, predicate family, two audit protocols (Drift Probe, Loop Budget Test), and a failure record. No cross-system superiority is claimed. Includes both English and Chinese versions.

Authors

Publication Details

Journal
Zenodo (CERN European Organization for Nuclear Research)
Published
2026-10-04
DOI
https://doi.org/10.5281/zenodo.23144014
Primary Topic
Artificial Intelligence Applications
Type
preprint
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
preprint

Stopping as a Database Property: Reason-Bearing Termination for Long-Running LLM Agents

Baofeng Zhao
Zenodo (CERN European Organization for Nuclear Research)
Artificial Intelligence Applications
preprint

Stopping as a Database Property: Reason-Bearing Termination for Long-Running LLM Agents

Baofeng Zhao
preprint en

Abstract

Loop termination should be a property of the data layer: when the substrate owns resource and intent state, stopping becomes a reason-bearing, auditable predicate rather than an arbitrary step cap or an unreliable model decision. We document this in a self-built agent research testbed. The substrate's signal ledger holds 48,061 self-monitoring events over 45 days; loop-related entries reference round counts up to 212 (208 distinct values). A mutual-critique relay loop emerged on its own and converged naturally at 29 rounds, seven weeks before any predicate formalized that behavior. A self-narration layer grew to 12,102 entries. On the official ARC-AGI-3 benchmark: 28.57 (3/6 levels, 31 steps) and 0.15 on Kaggle, operating under a self-imposed 82-step budget. We crystallize the mechanisms into four data-computable termination predicates, and document a recurring dissociation: convergence control and converged-to content are separable—a relay loop faithfully amplifies upstream wiring faults. Contributions: placement taxonomy, predicate family, two audit protocols (Drift Probe, Loop Budget Test), and a failure record. No cross-system superiority is claimed. Includes both English and Chinese versions.

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.