OpenAI preprint, September 2026 · §5. Switching leaves and completing the proof
§5 (proof of Proposition 5.1): weighted deficit:Finite incidence families with total leaf weight one and summed row
budgets at most
A good leaf from weighted deficit bounds.
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 weighted deficit lemma concerns detector losses distributed over a finite family of leaves. Let and be finite sets of vertices and leaves, respectively, and let be an incidence relation. The incidence set of a vertex is the set given by .
Let be a real detector value for and , and let and be real leaf weights and vertex budgets. The path deficit of a leaf is the sum of its incident detector losses: (1)✦
A bad leaf is a leaf with an incident detector value at most one half. The set of bad leaves is the subset of given by (2)The bad leaf mass is the total weight of the set in (2): (3)✦
The Claim
Theorem 1. Suppose that the leaf weights are nonnegative and sum to one, and that whenever . Suppose also that the row losses and total budget satisfy (4)Then the path deficits and bad leaf mass defined in (1) and (3) satisfy (5)✦
In the paper, the result reads: the block estimate gives total weighted path deficit at most and total weight at most for leaves with any path detector at most , so, since the leaf weights sum to one, some leaf has every path detector strictly greater than .
What Is Checked
The claim (5) is stated over an arbitrary incidence relation between finite sets, with real detector values and normalized nonnegative leaf weights, under the incident detector bound and the budget estimates (4).