← Moser's Worm Problem / Rounds / Round 12
Round 11's handoff called for recomputing all interval boxes using Arb/python-flint ball arithmetic, but this round's environment has no python-flint, Arb, or Sage, and no installable version was available from the package repositories. So directed-interval arithmetic via mpmath.iv is used as an independent fallback verification instead, without claiming to have completed an Arb replay. Both of the most critical differences replay successfully and independently, and are strictly positive: s₁₂₀-s₂₇₀∈[1.585e-9,1.692e-9], smooth candidate minus event control ∈[on the order of 1.058e-5]. All 19 closed-form boundaries are confirmed (19/19), and all 12 stationary-point root boxes pass interval Newton (12/12). But of the 18 internal boundaries, only 5 can be directly signed by mpmath.iv; for the rest, natural interval expansion near atanh(±1) (dependency inflation) is too wide to independently reproduce Round 11's narrow boxes — this is not a counterexample, Round 11's closed-form error bounds still stand, they simply were not independently replayed by this round's fallback library. The 579 derivative leaf boxes originally planned for one-by-one replay also could not be completed due to runtime limits. This round explicitly records “not completed” rather than writing it up as “passed.”
Relationship to other packages, stated as far as possible in the document's own words, not my interpretation.
“This round did not write ‘not completed’ up as ‘passed.’” — from Section 6, “What Remains Incomplete,” of this package's main document, an honest record of the failure to independently replay the 579 derivative leaf boxes one by one.
Loading…
Loading…
Loading…
Loading…
sha256 08b394e1ed9165d0e97973717a8cb4b81500478c28a7b589c6473075f6ae961a