Structural Retention Index (SRI): A Collapse Index Extension for Orthogonal Stability Assessment
Why this work is in the frame
A frame that forgets how it found something cannot be audited. These are the routes that admitted this work.
Bibliographic record
Abstract
This paper introduces the Structural Retention Index (SRI), a metric that quantifies how well AI systems preserve internal decision structure under perturbation. SRI measures stability dimensions orthogonal to the Collapse Index (CI), providing comprehensive dual-signal failure detection. Key Results: Perfect complementarity: CI + SRI = 1.000 (inverse measures of the same stability phenomenon) Equal discriminative power: Both achieve AUC=0.874 for error detection Vastly outperforms confidence alone (AUC=0.171) Identifies hidden instability: 20 Type II cases (4.0%) show internal confidence shifts without visible label flips Validation:AG News 4-class text classification (500 base examples × 4 variants = 2,000 predictions). Reproducible dataset, generation pipeline, and validation metrics publicly available. Methodology:SRI computation remains proprietary to prevent adversarial optimization. Validation outputs are independently verifiable. This approach balances scientific transparency with IP protection, consistent with industry practices for evaluation frameworks. Reproducibility: Dataset: agnews_ci_sri_demo.csv (2,000 rows) Generation script: generate_agnews_demo.py (MIT license) Validation script: validate_metrics.py (MIT license) GitHub: https://github.com/collapseindex/ci-sri License: CC BY-NC-ND 4.0 (Attribution-NonCommercial-NoDerivatives) Contact: ask@collapseindex.org | https://collapseindex.org
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.002 | 0.002 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.001 |
| Science and technology studies | 0.003 | 0.000 |
| Scholarly communication | 0.002 | 0.001 |
| Open science | 0.003 | 0.012 |
| Research integrity | 0.000 | 0.001 |
| Insufficient payload (model declined to judge) | 0.001 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from it