Code and schemas for a provenance-aware launchpad study: four-chain dataset construction, identification checks, real applications, calibrated experiments, and paper-artifact reproduction.
Bulk tables are not stored in this Git tree. The public dataset is kl41r3/web3ai4io-multichain-launchpad on Hugging Face.
uv sync --frozen
make reproduce
This regenerates empirical tables and figures from archived summaries, writes the artifact manifest and checksums, and checks manuscript numbers. It does not re-query chain data or call a model API.
| Command | Output |
|---|---|
make reproduce |
archived summaries, tables, figures, checksums, tests |
make figures |
empirical charts in ../paper/figs |
make tables |
tab_data_scope.tex and tab_claim_evidence.tex |
make paper |
compile the adjacent manuscript |
make verify |
files, checksums, sample counts, identity scan |
make all |
reproduce then compile |
See REPRODUCIBILITY.md and DATA_CARD.md.
dataset/ builders, schemas, event layer, release manifests
identification/ event registry, design checks, S1–S5 experiments
application/ PumpSwap applications, prompts, and plots
reproduction/ paper tables, figures, manifest, and tests
data/ local source pointers; generated tables are gitignored
Paper source lives at ../paper/ and is not part of this repository.
With local source bundles present:
dataset/.venv/bin/python dataset/scripts/build_solana_core.py
dataset/.venv/bin/python dataset/scripts/build_crosschain_core.py
dataset/.venv/bin/python dataset/scripts/build_events.py
PYTHONPATH=dataset/src dataset/.venv/bin/python -m pytest dataset/tests
Canonical Parquet output is written under data/canonical/. See dataset/README.md.