Supplementary materials for: Dynamic Reliability of Decision-Making under Measurement Channel Distortions: A Closed-Loop Model and Validation on the EvenFlow Dataset
Annotatsiya
Supplementary materials (revision 6) for the article "Dynamic Reliability of Decision-Making under Measurement Channel Distortions: A Closed-Loop Model and Validation on the EvenFlow Dataset". Revision 6 incorporates corrections made in response to peer review: goal-reaching logic, two error measures in the closed loop (D_local and D_ref), dimensionless decision-margin geometry, a corrected density threshold, a unified noise channel with bias as a separate variant, grouped bootstrap by source recordings with explicit Benjamini-Hochberg families, and a held-out check. The archive contains the vectorized agent (code/agent6.py) and analysis scripts, the revision 5 code with a validation script that reproduces revision 5 results, all raw results (results/, including report.json with every number in the article's tables), and Figures 1-8 in grayscale (figures_bw/). The EvenFlow dataset (Standard Cognition, https://huggingface.co/datasets/standard-cognition/EvenFlow, revision 307a024e) is not included; see README.md for reproduction instructions. Documentation and code comments are in Russian.
Hali tarjima qilinmagan