datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
unclutching-corpus-v1
unclutching-corpus-v1
Continued-pretraining corpus on the topic of Unclutching as articulated by SPH Bhagavan Sri Nithyananda Paramashivam.
Format
JSONL with a single field per row:
{"text": "..."}
Stats
Rows: 194
Source: synthesized long-form expository passages (synth-cpt-cli)
Use
Plain {"text": ...} shape suitable for masked / causal-LM continued pretraining.
unclutching-corpus-v2
unclutching-corpus-v2
Mixed continued-pretraining corpus on Unclutching combining synthesized expository passages with deduplicated excerpts from the Kailasa wiki book corpus.
Format
JSONL. Every row has text and source_dataset; rows from the wiki also carry origin metadata.
{"text": "...", "source_dataset": "synth-cpt"}
{"text": "...", "source_dataset": "kailasa-wiki", "corpus": "book", "source": "...", "id": "..."}
Composition
source_dataset
rows… See the full description on the dataset page: https://huggingface.co/datasets/Publishing/unclutching-corpus-v2.unclutching-corpus-v3
Unclutching Corpus v3
English-only, quality-filtered excerpts discussing the meditation concept of unclutching,
extracted from books and transcripts, de-duplicated, and cleaned.
Stats
Records: 1,012
Total characters: 679,435
Avg chars per record: 671
Schema
Single column:
{ "text": "string" }
Filtering pipeline
Built from a 3,766-row deduplicated source corpus by:
Dropping non-English sources (lang-tagged translations, non-Latin script).
Dropping… See the full description on the dataset page: https://huggingface.co/datasets/Publishing/unclutching-corpus-v3.unclutching-corpus-v4
Unclutching Corpus v4 — Domain/General Mix
A continued-pretraining mix combining the curated unclutching domain corpus with
general educational web text from FineWeb-Edu, at a roughly 25/75 domain/general ratio
by row.
Composition
Split
Source
Rows
Share
Domain
Publishing/unclutching-corpus-v3
1,012
~25%
General
HuggingFaceFW/fineweb-edu (sample-10BT, first 3,036 rows ≥200 chars)
3,036
~75%
Total
4,048
100%
Mix is shuffled with seed 42.… See the full description on the dataset page: https://huggingface.co/datasets/Publishing/unclutching-corpus-v4.core_concepts_100