What are the eight distinct definitions of 'model collapse' in Schaeffer et al. 2025 (arXiv:2503.03150), and does the full text cite Bartlett?

claim-model-collapse-literature-has-eight-conflicting-definitions rests on a Tier-1, quote-verified top-line count — Schaeffer, Kazdan, Arulandu & Koyejo (arXiv:2503.03150) state their survey found "eight distinct and at times conflicting definitions of model collapse." But the specific enumeration of those eight is flagged [unverified-mechanism -- needs primary]: the named sub-definitions (collapsing variance, tail loss, mode disappearance/entanglement, hallucinated-data appearance, …) were reconstructed from an aggregated search summary, not read from the primary text.

What would answer it.

• Read the full text of arXiv:2503.03150 (HTML or PDF) and extract the exact eight definitions verbatim, with the section/passage each is drawn from.
• While in the full text, check the bibliography and body for any citation of Bartlett, "serial reproduction," or cognitive-psychology reconstructive-memory work. The abstract contains none; confirming or refuting this for the full text sharpens observation-bartlett-typology-not-yet-operationalized-as-model-collapse-metrics.

Why it matters. The eight-way taxonomy is the AI-side half of the would-be Bartlett bridge in question-bartlett-distortion-typology-maps-to-model-collapse-failure-modes. Getting the exact definitions is a precondition for testing whether any of them line up one-to-one with assimilation, leveling, or sharpening. It also feeds a possible standalone note on definitional fragmentation in the model-collapse literature.