back to table · edit · history · where entries came from · files · download
6446 bytes, as of the version from 2026-11-20 03:35 (current). Recorded here, not run.
# Argument selection for T770
This is the scalar cyclotomic pairing B_p(P_i,P_j), not a collection of height
functions or a regulator determinant. Parameter order is (N,c4,c6,p,i,j).
All six parameters have type Z. There are no unassessed continuous couplings.
The natural family consists of rational elliptic curves with the specified
integral minimal model and Neron differential, good ordinary primes p>=5,
and indexed rational points. The finite source catalogue is an exploratory
arithmetic reference selection, not a complete catalogue of this family.
## Curve identity: N, c4 and c6
N is the conductor. c4 and c6 are invariants of the listed reduced global
minimal model, not independently sampled integers. They jointly identify the
curve; N is useful redundancy. The exact five coefficients and rational points
are in the mathematical document and points.txt. Models are never reconstructed
by guessing from rounded invariants or from mutable generator searches.
The source is Cremona ecdata commit 25cec5ecfec8b9f016eb1631ac633194c2bed39f,
allgens/allgens.00000-09999, SHA-256
d4200efe89ce8cb7874040711878fa11f04468e3181e9b8e43fd0f84ae0f1175.
The complete selected input is points.txt, SHA-256
049273399057ba95739695c3a848bc2bc1791bee35a4e5d90c8048e11471fc21.
The generator freezes those exact records; input_check.py independently
reconstructs their selection from the full pinned source and checks the
companion allcurves models/ranks and the conductor-10000 endpoint.
The selection takes the optimal representative in each class with source
rank one and N<=175, rank two and N<=2000, or rank at least three and N<=10000.
The last group contains just the rank-three curve 5077a1. Cremona's data
format/optimality documentation establishes curve number 1 in this range
except 990h3. That exception has source rank one and conductor 990, outside
the selected rank-one bound. The exception was explicitly checked.
These are noncanonical, notation-independent finite conductor budgets. They
preserve the wanted plan's useful rank-one overlap, provide broad rank-two
coverage, and retain a genuinely higher-rank anomalous-prime control. They
are not intrinsic cutoffs or an entry-count target. Increasing every conductor
bound at the expense of independent precision checks is not justified here.
No eligible member within the stated slice is pruned for difficulty or because
its value is less famous. Rank is an imported arithmetic selection attribute,
not a claim of an independent full Mordell-Weil certificate.
For each of N,c4,c6 a possible review category is reference_sample: the purpose
is recognition of point-pairing entries across bounded arithmetic curve
families; the bounds are the three rank-dependent conductor slices; the
coverage rationale is breadth in rank two plus low-conductor rank-one links
and a rank-three control. Noncanonical choices, notation independence and
checked source support are acknowledged. These parameters must be evaluated
jointly: the table is not a Cartesian box of N,c4,c6 values.
## Prime p
p>=5 is a prime of good ordinary reduction. The finite reference set is the
first four primes at least five: 5,7,11,13, then filtered separately for each
curve by exact good reduction and a_p!=0 mod p. This is a noncanonical,
notation-independent small-prime reference policy compatible with Harvey's
sigma implementation, the source controls, and the existing rank-one table.
It exercises ordinary and anomalous regimes without importing a free
continuous grid. p=13 also supplies held-out signed combination controls after
the sign convention is fixed by the p=5 documentation controls. No prime is
excluded because its height has negative valuation. There are 47 anomalous
objects. A possible review category is reference_sample with the stated
purpose, bounds and coverage rationale, not intrinsic.
## Point indices i and j
For each frozen row, the first source-rank points in the ecdata order are
P_1,...,P_r; the remaining source points are torsion controls only. Ordinary
projective [X:Y:Z] means affine (X/Z,Y/Z). All point equations are checked
exactly. No runtime E.gens(), generator negation, lattice reduction or basis
change redefines these identities. i and j are discrete structural indices
with 1<=i<=j<=r. Every upper-triangular matrix component, including all
diagonals, is included; symmetry supplies the lower triangle. There is no
arbitrary subgroup-coordinate grid in the numerical entries. A possible
review category is discrete_index for each index, with conditional bounds
coming from the exact point catalogue. Signed combination tests are evidence,
not extra table entries or a point-recognition uniqueness assertion.
## Enumeration and completeness
There are 60 rank-one curves, 97 rank-two curves and one rank-three curve.
The 541 eligible curve/prime objects contribute 193, 1032 and 24 entries,
respectively: 1249 total. Prime object counts are 134,137,128,142 for
5,7,11,13. input-evidence.json records an independent standard-library
finite-field enumeration; compute.py repeats eligibility in Sage.
The default generator, stored entry identities, Programs example and
complete-note must all describe this same slice. complete is no. The natural
parameter domains are broader than these finite budgets. The 49-entry soft
excess keeps all structural components of the coherent conductor slice;
there is no mathematical reason to discard arbitrary pairings or split ranks.
The actual block and largest-entry bytes are measured before upload.
## Sources and instruction provenance
Checked sources: Cremona ecdata documentation, sections TABLE ONE: CURVES and
TABLE TWO: GENERATORS (https://johncremona.github.io/ecdata/); pinned allgens
and companion allcurves files; Sage 10.9 padics.py, padic_height and
padic_height_pairing_matrix; Harvey arXiv:0708.3404v1, sections 2.2,4,5;
PARI Elliptic curves manual, ellpadicheight and ellpadics2. The source rank and
full-generator claims are imported evidence. Independent saturation is not
claimed, and determinants in evidence remain explicit-subgroup determinants.
Canonical skill archive: agents/runs/20261120T030205Z-build-skills/manifest.json.
Core SHA-256 b79b7fbc9fa3ce63f2d642cccdac12b0d214757ab01db66c03a2d84264a58e5f;
matching local/live bundle SHA-256
b5c3c7985b7c9e178b4899d083a9e3bc4284989dff76f05e5b0e52835d47eed2.
This is the builder's justification, not an independent argument-review verdict.