Supernatural_25k / README.md
gss1147's picture
Upload 2 files
6786b2c verified
# Supernatural_Archon_25k (Master Scholar)
**Supernatural_Archon_25k** is a 25,000-example dataset aimed at *master-scholar* depth on **supernatural topics as cultural, historical, and analytical material**.
It is designed for:
- Comparative mythology & folklore (motifs, functions, quant methods)
- Esoteric history & historiography (grimoires, alchemy, hermeticism) as **historical literature**
- Comparative religion & hermeneutics
- Scientific evaluation of paranormal claims (protocols, statistics, cognitive bias)
- Cognitive science of anomalous experiences (sleep states, agency detection, coincidence)
- Ritual studies & symbolic interpretation (semiotics, meaning, context)
- Fictional worldbuilding systems for supernatural settings (explicitly labeled as fiction)
- Ethics and safety around publishing supernatural-themed content
This dataset is **synthetic** and uses a **single consistent schema** across all records.
## Files
- `supernatural_archon_25k.jsonl` — 25,000 records (single schema).
- `dataset_card.md` — Hugging Face dataset metadata.
## Single schema (consistent across all records)
Each JSONL line contains:
- `id` (string): unique identifier
- `prompt` (string): task prompt (plain text; math included where relevant)
- `answer` (string): concise master-scholar answer
- `domain` (string): one of
- `mythology_folklore`, `esoteric_history`, `comparative_religion`,
`paranormal_investigation`, `cognitive_anomalies`, `ritual_literature`,
`occult_symbolism`, `worldbuilding_systems`, `ethics_safety`, `safety`
- `subdomain` (string): e.g., `motif_analysis`, `historiography`, `protocols`,
`coincidence`, `ritual_theory`, `correspondences`, `system_design`, etc.
- `task_type` (string): `compute` | `explain` | `derive` | `design` | `critique` | `interpret`
- `level` (string): fixed as `master_scholar`
- `verification.method` (string): `formula` | `closed_form` | `constructed` | `manual_review_recommended` | etc.
- `safety` (string): `allowed` or `refuse`
- `license` (string): `apache-2.0`
- `source` (string)
- `seed` (int)
## Domain composition
{
"mythology_folklore": 5000,
"esoteric_history": 3500,
"comparative_religion": 2500,
"paranormal_investigation": 3000,
"cognitive_anomalies": 2500,
"ritual_literature": 2500,
"occult_symbolism": 2000,
"worldbuilding_systems": 2000,
"ethics_safety": 1500,
"safety": 500
}
## Safety and responsible use
- The dataset **does not** provide instructions to harm, coerce, or target real people.
- A subset of examples are **refusals** for harmful requests (e.g., curses targeting individuals, coercion, or evasion).
- Where “ritual” language appears, it is framed as **literary/fictional** or as **historical/anthropological description**, not as a claim of real-world efficacy.
## Recommended use
- **SFT / instruction tuning:** train `prompt -> answer`.
- **Compute-only evaluation:** filter by `task_type == "compute"` and `verification.method` in
`{"formula","closed_form","constructed"}`.
- **Balanced modeling:** mix `paranormal_investigation` + `cognitive_anomalies` with `mythology_folklore` and `esoteric_history`
to build systems that are both culturally literate and statistically disciplined.
## License
Apache-2.0