Loading
Life sciences · Preprint
arXiv · September 10, 2026
Raises a question worth testing. It does not answer one.
This preprint reports a controlled computational study in which recursive training of 13 small language models (1–4B parameters) across five generations, with varying market concentration (including 90% oligarch share), produced collapse dynamics that were largely insensitive to market structure. The primary finding is an invariance: the speed and destination of model collapse were driven chiefly by the composition and susceptibility of the pool contributors, not by concentration or oligarch bias, within the tested parameter range.
Controlled computational simulation with multiple experimental arms. 13 open language models (1–4B parameters), arranged in ecosystems of 3 to 13 models per arm, with an injected probe to achieve 90% concentration in some arms.. Intervention: Recursive retraining on shared pool output; market concentration manipulation (up to 90% oligarch share); injected bias probe; partial human text substitution in pool.. Compared with: Varied ecosystem compositions and market share distributions; human text replacement vs. model-only pool..
Making market share more unequal barely changed the speed of collapse; endpoints shifted by only a few percent of common drift across all arms Swapping members of a K=3 ecosystem with fixed shares changed five-generation drift by 2.8×, indicating composition matters more than inequality A share-weighted susceptibility index explained speed differences across nineteen arms with R² = 0.68
Safety was not reported in the material analysed. Check the source before drawing any conclusion about harm.
The source did not state who this applies to in practice.
This is an exploratory computational study of a mechanistic phenomenon (model collapse in multi-model ecosystems) using controlled simulation, with no clinical, regulatory, or real-world validation endpoint and findings that raise questions rather than answering settled ones.
As stated by the source record.
Quoted from the source exactly as published.
Graded across the dimensions that decide whether you should act, each from what the source actually supports. There is no single score, and where a dimension was not assessed it says so.
AI-generated text is flowing back into the training corpora of the next generation of models. Recursive training on it drives model collapse, and recent work extends the setting to many models feeding one another -- but almost always with the market split evenly, while real generative AI is an oligopoly. Concentration raises two worries: fewer, more uniform sources may make collapse faster, and later models may be dragged toward the oligarch's output. We test both in controlled ecosystems: 13 open 1--4B models form natural ecosystems of 3 to 13 players, plus an injected probe that pushes the top share to 90%; each generation, every model's output is mixed into a shared pool by market share and every model is retrained on that pool from clean base weights, for five generations. Yet within the range we test, neither worry materializes; what emerges instead is an invariance. Making the split more unequal barely changes the speed of collapse. Destinations move even less: the share and identity knobs shift five-generation endpoints by only a few percent of the drift common to all arms -- the ecosystems collapse to nearly the same place. An extreme share paired with the strongest injected bias still does not guarantee steering, and the topic shifts it does produce leave only a faint trace on the ruler that measures collapse. What sets the speed is who supplies the pool and how readily those suppliers are carried along: with every share held fixed, swapping the members of a K=3 ecosystem changes five-generation drift by 2.8x; a share-weighted index of each member's susceptibility explains the speed differences across nineteen arms with R^2 = 0.68; and replacing half the pool with human text roughly halves drift without changing its course. Within the tested range, concentration sets neither the destination nor the pace of collapse; the pace follows whose text fills the pool.
Taken from the source record, never inferred. Follow any of these and new work involving them reaches your briefing.