kaal:claim:7598518-028

Claim. The pool's entire exposure to real data is 192 rows, below the 200 real rows the ten percent arm receives and far below the roughly 1,990 distinct real rows it trains on over the run.

Type. design · Support. stated

Holds when

• The keyed record is stratified to teach the pool who judges validity well on which tag, and nothing about where the rare components lie beyond a bounded tag leakage.
• The source's text layer drops an exponent inline, so the quote renders the formulas as "in one 50 generation" and "N(1 - 0. 9 )". The claim asserts only the three row counts, which the quote states in figures.

Source quote

Those KG = 192 rows are the pool's entire exposure to real data, below the ⌊0. 1N⌋ = 200 real rows the ten percent arm receives in one 50 generation and far below the N(1 − 0. 9 ) ≈ 1, 990 distinct real rows it trains on over the run.

From. Wulf A. Kaal, Can Reputation Prevent Model Collapse? A Preregistered Test of Reputation-Weighted Data Filtering (2026). Full text.

Cite as. Wulf A. Kaal, Can Reputation Prevent Model Collapse? A Preregistered Test of Reputation-Weighted Data Filtering (2026). SSRN: https://ssrn.com/abstract=7598518

Verify. sha256 of source PDF cd22d148b2b55c2e27f087dcaa61cdf44ad7022b400b355c31a9cd7e43a31749 at https://wulfkaal.com/papers/7598518.pdf

Content hash. e16b6a6fbc3d0dc122a51a1489f0a2ccf410cbc1d484994267e53259000a5917