Saved in:
| Main Author: | |
|---|---|
| Format: | Recurso digital |
| Language: | |
| Published: |
Zenodo
2026
|
| Online Access: | https://doi.org/10.5281/zenodo.19058064 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
Table of Contents:
- <p>This dataset and experiment showcase the behavior of a controlled large language model (LLM) under a governance-layer framework originally validated in a synthetic harness. While the synthetic system exhibited fully stable (“GREEN”) outputs even under extreme stress, scaling to a full LLM introduces observable variability in the form of RED spikes, AMBER/BLUE transitional states, and extended GREEN stability zones. These patterns reveal emergent instabilities, cumulative drift, and dynamic recovery points, offering an empirical view of governance-layer performance at scale. The visualization provided highlights these zones, making the underlying stability dynamics immediately interpretable. This work demonstrates that RED spikes do not indicate failure but rather signify active governance response to high-dimensional LLM variability, validating the robustness and adaptive behavior of the system under realistic operating conditions.</p>