Run Results

Result tables summarize bounded runs. They report what a configured search inspected; they do not prove or verify a conjecture.

Repository Ledger Viewer

How To Read This Page

near-miss

A ranking score inside one strategy, not a probability.

unrefuted

No counterexample found within this run's budget.

ledger

The candidate-level audit trail behind a summary.

Checking for generated results…

No Generated Results Yet

This page shows real search output once src/results.json has been generated. Run the reproduction command below, then reload this page.

Run Summary

Verified Run Artifacts

These artifacts bind each public engine demo to a seed, bound, commit, claim level, and the environment the run was built in. Expand Environment provenance on any card for the interpreter, platform, and resolved dependency versions; a card whose build environment could not resolve every declared dependency is labelled incomplete rather than presented as a full snapshot. They are reproducibility records, not mathematical proofs.

CollatzX Near-Miss Candidates

Rows are ranked by the Collatz near-miss score (0.70 × risk_score + 0.30 × excursion_bonus, see derivation). A listed candidate may still converge normally; high score only means the strategy considered it worth surfacing.

#	Candidate	Near-Miss	Stopping Time	Peak Value	Strategy

bounded runnot a proof

GoldbachX Deficit Candidates

Rows are ranked by the Goldbach near-miss score (0.85 × deficit + 0.15 × structural_hardness, see derivation). A deficit is a search signal, not a calibrated probability of failure.

#	Candidate	Predicted G(n)	Actual G(n)	Near-Miss	Strategy

bounded runnot a proof

RiemannX Diagnostics

RiemannX output is numerical diagnostic data. Publish precision, library versions, and code revision before treating any table as research evidence.

Diagnostic	Meaning	Limit
zero_deviation_scan	Measures numerical deviation from the critical line.	Precision and method dependent.
keiper_li_coefficients	Computes coefficient signals for inspection.	Numerical, not certified.
prime_counting	Compares exact capped counts to simple approximations.	Browser demo is capped.

Reproduce A Run

Generate the public verified-run artifact bundle with:

python -m codebase.cli run all ^
  --seed 42 ^
  --output-json src/verified-runs.json

Generate a deeper candidate summary and ledger with:

python -m codebase.cli falsify ^
  --budget 500 ^
  --seed 42 ^
  --target both ^
  --top-k 25 ^
  --save-ledger results/ledger.jsonl ^
  --output-json src/results.json

On Unix shells, replace `^` line continuations with `\`. Or run ./scripts/generate-results.sh (.ps1 on Windows).