docs/UPLOAD_GUIDE.md

Upload to Hugging Face

Suggested repository slug: QIRAN-EIDOLITH-v4
Suggested type: Dataset
Suggested full title: QIRĀN–EIDOLITH v4: Hereditary Search, Neutral-Corridor Compilation, and Reusable Evidence for Recursive Self-Improvement

This is a research-artifact repository with three small browseable metadata configurations. It contains source code, experimental records, and a manuscript. It does not contain pretrained model weights. Uploading is a separate action; preparing this package has not created or published a Hub repository.

Web interface

1. Create a dataset repository in your own namespace, using the suggested slug or another name you prefer. Select visibility deliberately.
2. Extract the release ZIP locally.
3. Upload the contents of QIRAN_EIDOLITH_v4_HuggingFace/, so README.md, research/, data/, docs/, and tools/ are at repository root. Uploading only the ZIP will not expose the card or structured indexes.
4. Keep README.md as the dataset card. Its YAML selects only the three curated JSONL indexes, avoiding accidental merging of heterogeneous research outputs.
5. Confirm that the manuscript, local evidence links, and claims/metrics/artifacts configurations are visible after upload.

The README can also be copied directly into the card editor if the files are uploaded separately. Leave the mixed-license metadata intact unless you deliberately change the applicable licenses with authority to do so.

Windows: double-click uploader

1. Extract the complete release ZIP.
2. Double-click UPLOAD_TO_HUGGINGFACE.bat in the extracted QIRAN_EIDOLITH_v4_HuggingFace folder.
3. The launcher detects Python 3.10+ and installs huggingface_hub if missing. If Python is absent it gives installation instructions.
4. The wizard uses a saved Hugging Face login or requests a write token in a hidden prompt. It suggests a repository in your authenticated account; press Enter to accept it or enter another permitted namespace/repository.
5. For a new repository, choose public or private. For an existing one, confirm updating its same-name files; its visibility is preserved.
6. Wait for the upload-complete link. The window remains open so errors stay readable.

An uploader add-on ZIP is also provided for people who already extracted the original release. Extract it into the same parent directory as your existing QIRAN_EIDOLITH_v4_HuggingFace folder and allow its updated files to replace the originals. The add-on requires the original research files; use the full corrected ZIP if unsure.

Offline verification, without login or network access:

python UPLOAD_TO_HUGGINGFACE.py --dry-run

Direct command with an explicit target:

python UPLOAD_TO_HUGGINGFACE.py YOUR_NAMESPACE/QIRAN-EIDOLITH-v4 --public

Add --allow-existing when intentionally updating an existing dataset from the command line. The uploader selects only checksum-listed files; it does not delete unrelated remote files or store your prompted token in the release.

Command-line upload

Install the Hub client and authenticate locally:

python -m pip install -U huggingface_hub
hf auth login

Do not put access tokens in the repository, an issue, or a chat message.

From the extracted release root, the included helper defaults to a private new repository:

python tools/upload_hf.py YOUR_NAMESPACE/QIRAN-EIDOLITH-v4

To create a public new repository explicitly:

python tools/upload_hf.py YOUR_NAMESPACE/QIRAN-EIDOLITH-v4 --public

The helper verifies local integrity, prints the target and visibility, creates a new dataset repository, and uploads the complete prepared tree in one commit. It refuses an existing repository unless --allow-existing is supplied; existing visibility is not changed. Same-name remote files may be replaced on an explicitly allowed update. It does not delete unrelated remote files.

Equivalent CLI steps for a new private repository:

hf repos create YOUR_NAMESPACE/QIRAN-EIDOLITH-v4 --type dataset --private
hf upload YOUR_NAMESPACE/QIRAN-EIDOLITH-v4 . . --repo-type dataset --commit-message "Release QIRAN-EIDOLITH v4 research artifact"

The helper is preferred because it filters local caches and newly generated files. A direct upload of . may include later local work. Review the upload selection before using the CLI on a modified tree.

Description for the repository creation form

QIRĀN–EIDOLITH v4 is a complete finite-model research release on continuation-safe self-modification: hereditary program equivalence, exact neutral-search compilation, state–history recovery, and reusable adaptive evidence. Includes a 125-page manuscript, proof sources, code, raw synthetic experiments, negative controls, and reproduction tools. General recursive superiority remains unestablished.

Useful topic tags

recursive-self-improvement, self-modifying-systems, program-semantics, formal-methods, adaptive-validation, exact-arithmetic, markov-chains, research-artifact, synthetic-data.

Official documentation

• Dataset cards and YAML metadata
• Dataset configurations
• Hugging Face CLI
• Upload files and folders

The examples follow documentation checked on 17 October 2026. Use the installed command’s --help if a later client changes its options.