The Curse of Recursion / Model Collapse
Updated 2026-08-05
INTRODUCTION
English translation pending.
CORE DEFINITION
Model collapse, described by Shumailov and colleagues, occurs when generative models are trained predominantly on data produced by earlier models. Errors and distributional narrowing compound across generations, producing irreversible loss of diversity and fidelity. The condition is recursion on synthetic output; training on fresh human data prevents the decay.
SCAFFOLDING EFFECT
Reduce cognitive load
- Source hygiene: Read primary material rather than summaries of summaries. - Chain tracing: Ask how many times information has been reprocessed before it reached you. - Degradation watch: Treat repeated restatement as a sign that content is thinning out.
Anchor fast decisions
Each generation samples from a distribution that has already lost its rare cases, so the tails are never regenerated. Errors are not corrected, because nothing in the loop refers back to the original data. Iterating compounds both effects, and the loss becomes irreversible once the source distribution has been dropped.
MINIMUM ACTION
In progress 0/1Practice this model in one real situation:
account_treeGenealogyexpand_more
menu_bookReferencesexpand_more
Source support: Explicit
- en.wikipedia.orghttps://en.wikipedia.org/wiki/Model_collapseverified
PRIVATE NOTES · Only visible to you
SAVED Q&A
ENTRY Q&A · Private saving available
Ask with a clear boundary
thinkingmodels answers from published entry context only.
Your question is sent to thinkingmodels. The answer uses public entry context only.
RELATED MODELS