The Collapse Index (CI) CrackTest: Morphology-Aligned Perturbation Testing Reveals Systematic Collapse Inheritance in Frontier Language Models
Bibliographic record
Abstract
This preprint introduces the Collapse Index (CI) CrackTest, a morphology-aligned perturbation framework for evaluating robustness and collapse inheritance in large language models (LLMs). The study demonstrates that CI CrackTest, a bounded and lightweight perturbation protocol originally developed for brittleness diagnostics, can quantify systematic error propagation across morphologically-related variants in controlled classification tasks. Using a 186-variant perturbation suite spanning eight morphological families (lexical, syntactic, ambiguity, semantic, compression, noise, boundary, contrastive), the evaluation analyzes collapse inheritance, family-specific brittleness, and confidence behavior under perturbation. Across three frontier models (GPT-4o, Claude Haiku 4.5, Gemini 2.5 Flash), the framework identifies consistent robustness signatures, including 11–16% collapse inheritance rates, 9–28% semantic brittleness, and 0.10–0.17 confidence masking deltas, independent of architecture or training regime. The framework is presented as a behavioral diagnostic tool for robustness analysis. CI CrackTest does not expose internal perturbation heuristics, variant-generation mechanisms, or scoring systems. All reported findings are based solely on externally observable model outputs under controlled morphological perturbation.All internal algorithms, classification mechanisms, and inference procedures remain proprietary to Collapse Index Labs. Project page: https://collapseindex.orgLicensed under CC BY-NC-ND 4.0.
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.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.004 | 0.024 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.001 | 0.000 |
| Science and technology studies | 0.001 | 0.002 |
| Scholarly communication | 0.002 | 0.002 |
| Open science | 0.001 | 0.003 |
| Research integrity | 0.001 | 0.002 |
| Insufficient payload (model declined to judge) | 0.003 | 0.001 |
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 itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".