Verified Outputs

upg-strings

Verified ML-guided search for rare Calabi-Yau targets (string landscape)

We build ML-guided search tools that drastically reduce the cost of finding rare Calabi-Yau geometries in large string-theory datasets, with full verification and reproducibility.

We achieve perfect precision and non-trivial recall in ML-guided search for rare targets, with sub-second runtime.

Run the Demo

Click the button below to run a live ML-guided search on 5,000 Calabi-Yau manifolds. Results will appear in ~5-10 seconds.

For command-line usage:

python cy_search_real.py

What It Does

1

Download Dataset

Fetch Calabi-Yau dataset with checksum verification

2

Train Baseline Model

Train ML classifier with fixed random seed for reproducibility

3

Rank Candidates

Score and rank candidates by predicted likelihood

4

Verify Top-k

Validate top results against ground truth labels

5

Export Artifacts

Generate CSV/JSON outputs with reproducibility metadata

For Scientists

Use upg-strings as a reproducible filter for rare geometries before deeper analysis.

Outputs

results_topk.csv

Top-ranked candidates with scores and verification status

metrics.json

Performance metrics and timing data

repro.md

Complete reproducibility report with environment details

Metrics

Precision@k

Fraction of top-k predictions that are verified correct

Recall@k

Fraction of all true targets found in top-k results

Time-to-First-Hit

How quickly the first verified target is discovered

Links

Disclaimer: This tool accelerates computational search and verification. It does not claim to 'solve string theory'.