OpenAI preprint, September 2026 · §E. Optimization on common log canonical slices
§E.4 (setting): local rounding regions:Exponent simplices and coefficientwise admitting regions for Laurent monomials
Rounded conditions admit strict linear descriptions.
Hover over a blue or green term in the text, or a node in the graph, to find it in the other. Select it to read its
declaration both in Lean and in words.
The Setting
The exponents assigned to summand ideals determine coefficientwise rounding conditions for a Laurent monomial. Fix a finite index set of summands and an index set of coefficient conditions. Let be the total exponent, let be coefficient data, and let be the Laurent monomial exponents, for and . An exponent vector is a real vector indexed by .
The Claim
Definition 1. The exponent simplex of mass is the subset defined by (1): (1)✦The admitting region for the coefficient data is the subset defined by (2): (2)✦For an integer and real numbers , the rounding criterion is (3): (3)✦For , membership in the admitting region has the strict linear description (4): (4)✦
In the paper, the result reads: resolving the summand ideals and their sum in the proof of convex domination yields divisors and , and with exponent denominator , on the exponent simplex , a fixed Laurent monomial with exponents belongs to the mixed rounded divisor sheaf exactly when for every , where are the coefficients of and are the relative canonical coefficients.
What Is Checked
The exponent simplex, admitting region and rounding criteria are stated for a finite set , any index set , arbitrary real , and integer .