datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
indicmarco-triplesharmonia-triples-rust-code-traversal
harmonia-triples-rust
Triples for source rust emitted by the ingest pipeline (current wave: v0.7). Schema: (s, p, o, src) with full provenance per ADR-0011. Pre-HHEC.
Provenance
Each parquet shard carries the full provenance chain per ADR-0011:
s, p, o, src columns (when this is a triples-stage dataset)
src = "<dataset>:<version>:<file>" for triples
Causal registry events recorded at causal_registry/master.jsonl chain
Architecture
Part of Harmonia… See the full description on the dataset page: https://huggingface.co/datasets/Hailstone-Technologies/harmonia-triples-rust-code-traversal.harmonia-triples-stackexchange-document-traversal
harmonia-triples-stackexchange-slice
Triples for source stackexchange-slice emitted by the ingest pipeline (current wave: v0.6). Schema: (s, p, o, src) with full provenance per ADR-0011. Pre-HHEC.
Provenance
Each parquet shard carries the full provenance chain per ADR-0011:
s, p, o, src columns (when this is a triples-stage dataset)
src = "<dataset>:<version>:<file>" for triples
Causal registry events recorded at causal_registry/master.jsonl chain… See the full description on the dataset page: https://huggingface.co/datasets/Hailstone-Technologies/harmonia-triples-stackexchange-document-traversal.opengloss-v2.0-retrieval-triples
Superseded by OpenGloss v2.1 (2026-09-07): 109,633 lexemes and 250,003 live senses — twice this release's coverage — plus a new opengloss-v2.1-inflections form→lemma lookup. v2.0 stays published for reproducibility.
OpenGloss v2.0 — Retrieval Triples
Training triples for an embedding model or reranker. Each row is a query, a positive passage from the sense the query was written for, and one negative. The hard negative is drawn from the graph by a priority-ordered fallback —… See the full description on the dataset page: https://huggingface.co/datasets/mjbommar/opengloss-v2.0-retrieval-triples.opengloss-v2.2-retrieval-triples
Superseded by OpenGloss v2.3 (2026-09-09): tier 6 adds ~12,000 named entities (people, places, organizations, works, events) with entity_type, Wikidata ids and alias_of links, and every proper noun in the release is now typed. v2.2 stays published for reproducibility.
OpenGloss v2.2 — Retrieval Triples
Training triples for an embedding model or reranker. Each row is a query, a positive passage from the sense the query was written for, and one negative. The hard negative is… See the full description on the dataset page: https://huggingface.co/datasets/mjbommar/opengloss-v2.2-retrieval-triples.opengloss-v2.1-retrieval-triples
Superseded by OpenGloss v2.2 (2026-09-08): 148,292 live lexemes and 288,304 senses — tier 5 closes the WordNet gap (38,100 entries imported from Princeton WordNet 3.0 and enriched), inflected-form headwords are folded onto their lemmas, and every inherited field carries a migrate provenance record. v2.1 stays published for reproducibility.
OpenGloss v2.1 — Retrieval Triples
Training triples for an embedding model or reranker. Each row is a query, a positive passage from the… See the full description on the dataset page: https://huggingface.co/datasets/mjbommar/opengloss-v2.1-retrieval-triples.opengloss-v2.3-retrieval-triples
Superseded by OpenGloss v2.4 (2026-09-25): every sense now has search queries, QA pairs and verified examples (v2.3 had them only for core and tier 2); level x register definitions and leveled contrasts and explanations are added; and the pretraining corpus no longer contains duplicate documents. v2.3 stays published for reproducibility.
OpenGloss v2.3 — Retrieval Triples
Training triples for an embedding model or reranker. Each row is a query, a positive passage from the… See the full description on the dataset page: https://huggingface.co/datasets/mjbommar/opengloss-v2.3-retrieval-triples.msmarco-embed-DenseOn-triples-v3
much1na/msmarco-embed-DenseOn-v4
Source Dataset: tomaarsen/msmarco-Qwen3-Reranker-0.6B
Embedding Model: lightonai/DenseOn
This is the inflated version with hard negatives + embeddings in every row. Has Normalized embeddings
medical-triplesarxiv-triples-filteredMedQuAD_Context_Question_Answer_Triples_TWO
Dataset Card for "MedQuAD_Context_Question_Answer_Triples_TWO"
More Information needed
msmarco-embed-DenseOn-triples-v2
much1na/msmarco-embed-DenseOn-v2
Source Dataset: tomaarsen/msmarco-Qwen3-Reranker-0.6B
Embedding Model: lightonai/DenseOn
This is the inflated version with hard negatives + embeddings in every row
MedQuAD_47441_Context_Question_Answer_Triples
Dataset Card for "MedQuAD_47441_Context_Question_Answer_Triples"
More Information needed
combined_triples_with_marginswikidata_rdf_massive_objects_EN-triples-and-sentences
A WIKIDATA based triples and sentence combination dataset for training on natural language relations.
Subset extraction
The RDF representation of a Wikidata item contains a redundancy, since it contains both the full statements and the "truthy" statements.
A subset that contains is not always necessary and being able to separate truthy from full statements lead to smaller subsets.
Similarly being able to taylor which part of Wikidata items (ie. Labels/descriptions, statements, and… See the full description on the dataset page: https://huggingface.co/datasets/creativeautomaton/wikidata_rdf_massive_objects_EN-triples-and-sentences.filtered_triples_with_marginsst_specter_train_triplessubject-triplesrelational-triples-1000_comlmsys-triples-dedupannotated_conversation_satisfaction_triples
Dataset Card for annotated_conversation_satisfaction_triples
This dataset has been created with Argilla. As shown in the sections below, this dataset can be loaded into your Argilla server as explained in Load with Argilla, or used directly with the datasets library in Load with datasets.
Using this dataset with Argilla
To load with Argilla, you'll just need to install Argilla as pip install argilla --upgrade and then use the following code:
import argilla as rg
ds =… See the full description on the dataset page: https://huggingface.co/datasets/acmc/annotated_conversation_satisfaction_triples.usc-knowledge-graph-triplesreranker-triples-multi
reranker-triples-multi
Mined hard-negative triples for cross-encoder reranker training, across 25 languages — (query, positive, negatives[5]), built from olaverse/qg-passages-multi.
Dataset Summary
For each (query, positive) pair, up to 5 hard negatives — passages that are semantically similar to the query but are not its true answer — mined via embedding similarity within a controlled rank window, with false-negative guards.
Data Fields
Field… See the full description on the dataset page: https://huggingface.co/datasets/olaverse/reranker-triples-multi.lmsys-triplesmondo_triples_ontologyrelational-triples-100_relation_listrelational-triplessubset_raw_query_triples_jan12anli_triplesMNLP_M2_RAG_retriever_triples