What was already known
BCHM proved finite generation for ordinary projective klt pairs. Generalized klt finite generation was known in dimension at most three and under specified extra positivity conditions; Totaro’s invariant ring and Brion’s extension are earlier ingredients.
AI involvement
Liu and Wang credit GPT-5.6 Sol Ultra, Fable 5 and Danus with obtaining the main result. They authored and polished the paper, using earlier human results as ingredients.
The source explicitly attributes the main construction to several AI systems, enough for central AI work, but gives no role-specific provenance that supports AI-led or autonomous attribution.
Importance
The example disproves unrestricted finite generation for projective klt generalized pairs.
Limitations
It does not challenge the ordinary klt theorem. The paper reports human verification without identifying the checker, checked steps or method; the roles of the three named AI systems are not separated.
Technical verification
Remark 1.5 reports later human verification, but neither a named checker, exact checked claim nor completed checking method and coverage are specified. Under the evidence rule this does not establish a Formal or Partial technical rating; it is a gap in documentation, not evidence of failure.
Human understanding
Default for missing attribution of a human explanation of the central AI-origin connection. The paper explains the construction, but says its main result was AI-obtained and only specifies later human verification and polishing.
Correction and publication history
The example disproves unrestricted finite generation for projective klt generalized pairs.
Read this published revision