CREAIM
CREAIM · RESEARCH CHALLENGES · 2026

Real Problems. Reproducible Results.

CREAIM documents local computational experiments and reproducibility audits against public optimization benchmarks, preserving successes, negative results and claim limits.

Sign in to view evidencePublic research overview

Status: independent benchmark research; official registration/submission not confirmed; no best-known record.

MEASURED LOCAL RESULTS · C2_10_4

Research progress

R2-02
10/10

Locally checked seeds

R2-05
3/3

Fixed-work repeat runs

R2-05
23/23

Local evidence integrity

These are owner-laptop local observations, not official competition acceptance or proof of MetaCore performance advantage.

RESEARCH TIMELINE

Transparent experiment timeline

R2-02LOCAL PASS

Real vehicle-routing trials

10 / 10 local independent checker PASS. Batch best 28 vehicles, 16,341.59; prior one-run result 28, 16,318.21. Reference BKS not beaten.

R2-04CAUSE HOLD

Same-seed inconsistency found

Seed 1 initially 28 vehicles, later 29. Same pinned input, version and 120s limit. Root cause not proven.

R2-05LOCAL PASS

Fixed-work repeatability

100 iterations × 3 identical ordered-route results, 42 vehicles each; 23 / 23 local evidence files re-verified.

R2-06PLANNED

Larger work-budget experiment

Future test only; no projected improvement or official submission.

XLTW · VRPTW

Routing benchmark with locally verified runs and explicit reproduction boundaries. Not submitted.

Detail & login

QOBLIB · MIS / LABS

Separate external pull-request validation track. Do not conflate validator results with merge or XLTW participation.

QOBLIB Evidence
CLAIM BOUNDARY · Local PyVRP baseline only; MetaCore and Q-Core not run. Historical same-seed mismatch remains unresolved. Local hash checks are not external attestation and raw automatic upload is not implemented.

CURATED PUBLIC SUMMARY · 2026-10-23 · NOT A LIVE DATABASE FEED