Hyukkyu/train-rlhn-arguana

RLHN ArguAna — Training, unified schema

The arguana subset of rlhn/rlhn-680K (RLHN: the BGE retrieval training collection with its false hard negatives relabelled by a GPT-4o-mini → GPT-4o cascade; Thakur et al., 2025), reshaped into the strict schema shared by every dataset in this collection.

Source	rlhn/rlhn-680K @ d5323fad7aea, subset arguana
Task	argument → counter-argument
Domain · languages	argument · eng
Queries / documents / qrels	2,508 / 7,713 / 3,155
Qrels per query	min 1 · mean 1.26 · max 4
Score values	2 ×3,155 (2: RLHN's original positive, 1: a hard negative RLHN relabelled as relevant)
Layout	queries · corpus · qrels · hard-negatives · teacher-scores, split train
Splits	corpus: train · hard-negatives: train · qrels: train · queries: train · teacher-scores: train
Hard negatives	sources: dataset, dense · 290,374 rows
Teacher scores	none yet (0 rows): jinaai/jina-reranker-v3.5 scores come next
Ids	sha1(text)[:20]; identical texts collapse to one document / query
License	cc-by-sa-4.0

Schema

config	columns	rules
queries	id: string, text: string	ids unique and non-empty; every query has ≥ 1 qrel
corpus	id: string, title: string, text: string	title is always present ("" when the source has none)
qrels	query-id: string, corpus-id: string, score: int32	referential integrity to both tables; no duplicate pairs; no floats
hard-negatives	query-id: string, corpus-id: string, rank: int32, source: string	one row per negative; (query-id, corpus-id, source) unique; never a labelled positive of the same query
teacher-scores	query-id: string, corpus-id: string, teacher: string, score: float32	one row per scored pair (positives included); a row means scored — never a placeholder

Files are Parquet, sorted by id, zstd-compressed, sharded at 500 MB. Every rule above is checked before publishing; provenance.json records the source revision, what changed, and the output file hashes. teacher-scores holds no rows yet; datasets cannot return a 0-example split, so read that file with pyarrow / pandas until it is filled.

What changed from the source

• subset: the rows of rlhn/rlhn-680K whose subset is arguana (ArguAna in the BGE training collection)
• positives: each query's positive passages; RLHN's original positive at grade 2, the hard negatives its LLM cascade relabelled as relevant at grade 1
• negatives: RLHN's remaining hard negatives in RLHN's order (rank 1, 2, …), source = dataset
• decontaminated (exact queries): 370 queries whose normalised text equals an evaluation query (the 23 test sets and the 6 dev sets) were removed
• decontaminated (overlap audit): 304 passages that equal or nearly copy an evaluation document some evaluation query judges relevant, and 303 queries that nearly copy an evaluation query (word 13-grams for passages, 8-grams for queries; at least half shared with one evaluation text) were removed, and with them 317 queries in total; near copies of evaluation-corpus documents that no evaluation query judges relevant were kept
• text: leading and trailing whitespace stripped
• ids re-keyed to sha1(text)[:20]: 0 documents and 870 queries collapsed into identical texts
• not published: the order scores the local conversion uses in place of teacher scores (RLHN has no teacher)

Hard negatives and teacher scores

Filled by the collection's annotation pipeline (annotation=jina35). Interim: the candidates are final, the teacher scores are still to come.

• Candidates: dense retrieval with jinaai/jina-embeddings-v5-text-small over this corpus to depth 1,000; 100 candidates per query drawn from the rank windows 1–30 (30), 31–100 (30), 101–300 (20), 301–1000 (20), the query's labelled positives excluded. rank is the dense rank; source is dense for a mined row and dataset for a negative the source labels itself.
• Teacher scores: none yet. teacher-scores holds 0 rows until the jinaai/jina-reranker-v3.5 scores (listwise, as in the other repositories) are filled in; datasets cannot return a 0-example split, so read that file with pyarrow / pandas meanwhile. The candidates stay.

configs	queries	hard negatives	teacher scores
hard-negatives · teacher-scores	2,508 (all)	290,374 (42,633 dataset, 247,741 dense)	0

Load it

from datasets import load_dataset
queries   = load_dataset("Hyukkyu/train-rlhn-arguana", "queries", split="train")
corpus    = load_dataset("Hyukkyu/train-rlhn-arguana", "corpus", split="train")
qrels     = load_dataset("Hyukkyu/train-rlhn-arguana", "qrels", split="train")
negatives = load_dataset("Hyukkyu/train-rlhn-arguana", "hard-negatives", split="train")

License and attribution

RLHN is released under CC BY-SA 4.0, and so is this copy (share-alike). The underlying ArguAna data carries its own terms. All credit belongs to the original authors; see rlhn/rlhn-680K. This repository is an independent repackaging. Please cite RLHN:

@misc{thakur2025relabel,
      title={Fixing Data That Hurts Performance: Cascading LLMs to Relabel Hard Negatives for Robust Information Retrieval},
      author={Nandan Thakur and Crystina Zhang and Xueguang Ma and Jimmy Lin},
      year={2025},
      eprint={2505.16967},
      archivePrefix={arXiv},
      primaryClass={cs.IR},
      url={https://arxiv.org/abs/2505.16967},
}