README.md

---
pretty_name: BEIR · FEVER
license: cc-by-nc-sa-3.0
language:
- en
multilinguality:
- monolingual
task_categories:
- text-retrieval
task_ids:
- document-retrieval
tags:
- beir
- mteb
- retrieval
- fact-checking
configs:
- config_name: corpus
  data_files:
  - split: test
    path: corpus/test-*.parquet
- config_name: hard-negatives
  data_files:
  - split: train
    path: hard-negatives/train-*.parquet
- config_name: judgments
  data_files:
  - split: train
    path: judgments/train-*.parquet
- config_name: qrels
  data_files:
  - split: test
    path: qrels/test-*.parquet
  - split: train
    path: qrels/train-*.parquet
  - split: dev
    path: qrels/dev-*.parquet
- config_name: queries
  data_files:
  - split: test
    path: queries/test-*.parquet
  - split: train
    path: queries/train-*.parquet
  - split: dev
    path: queries/dev-*.parquet
- config_name: teacher-scores
  data_files:
  - split: train
    path: teacher-scores/train-*.parquet
---
# FEVER — BEIR, unified schema

A normalised copy of the dataset behind the `mteb` task **`FEVER`**, one of the tasks of the [BEIR](https://github.com/beir-cellar/beir) benchmark as `mteb` defines it. Same queries, documents
and relevance judgements as the benchmark evaluates — reshaped into one strict schema shared by every dataset
in this collection.

|  |  |
|---|---|
| Source | [`mteb/fever`](https://huggingface.co/datasets/mteb/fever/tree/bea83ef9e8fb933d90a2f1d5515737465d613e12) @ `bea83ef9e8fb` (the revision pinned in `mteb`) |
| Domain · languages | fact-checking · eng |
| Queries / documents / qrels (all splits) | 123,142 / 5,416,568 / 156,101 |
| Qrels per query | min 1 · mean 1.191 · max 15 |
| Score values | 1 ×7,937 |
| Layout | `queries` · `corpus` · `qrels`, split `test`; `queries`/`qrels` also carry `train`, `dev` — one shared corpus; `hard-negatives` and `teacher-scores` for `train` |
| Splits | `hard-negatives`: train · `qrels`: train, dev, test · `queries`: train, dev, test · `teacher-scores`: train |
| Training extension | `hard-negatives` and `teacher-scores` for the `train` split are the owner's own mining and scoring ([details](#hard-negatives-and-teacher-scores)); `queries`/`qrels` `train` are the benchmark's training data |
| Hard negatives | sources: `dense` · 9,889,255 rows |
| Teacher scores | `jinaai/jina-reranker-v3.5` · 9,989,255 rows (positives included) |
| Judgments | `judgments`: `typesafe/jev-1.13.0` · 2,366,979 rows |
| License | `cc-by-nc-sa-3.0` |

## 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 mined 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 |
| `judgments` | `query-id: string`, `corpus-id: string`, `judge: string`, `role: string`, `p_yes: float32`, `round: int32` | one row per judged pair; `role` is `positive` (the training positive) or `candidate` (a mined candidate, never a labelled negative); `p_yes` in [0, 1]; `round` 0 the first request, 1.. top-ups |

Files are Parquet, sorted by id, zstd-compressed, sharded at 500 MB. Every rule above is enforced by a validator
before publishing; `provenance.json` records the source file hashes, what changed, and the output file hashes.

## What changed from the source

- **byte-preserved** all text — no whitespace, newline, or control-character normalisation
- **kept 3 qrels rows whose document is absent from the corpus** — a property of the source (the source's qrels reference document ids that its corpus does not contain). `mteb` evaluates the source as-is, and removing them would change the ideal DCG and break equivalence with the benchmark, so they stay; the referential-integrity rule below is relaxed for this dataset only
- `dev`: dropped 116,476 queries that have no qrels in this split (mteb does the same at load time)
- `test`: dropped 116,476 queries that have no qrels in this split (mteb does the same at load time)
- `train`: dropped 13,332 queries that have no qrels in this split (mteb does the same at load time)
- cast `qrels.score` double -> int32
- kept 31,703 documents whose text is identical to another document's — a property of the source; note that models embed such documents identically and retrieval ties among them are broken by corpus order
- renamed `_id` → `id`
- renamed source splits (`queries`←`queries/queries`, `corpus`←`corpus/corpus`) to `test`

## Hard negatives and teacher scores

Filled by the owner's annotation pipeline (`annotation=jina35`) for the `train` split of the
query set(s) below; queries without a labelled positive are left out.

- **Candidates**: dense retrieval with `jinaai/jina-embeddings-v5-text-small` over the full 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
  (those are kept for every query of the split, sampled or not).
- **Teacher**: `jinaai/jina-reranker-v3.5`, listwise: a query's positive and all of its candidates are scored together in one
  context of up to 32,768 tokens. `score` is the raw cosine score, one row per (query, positive) and per
  (query, candidate); a labelled negative that was also mined is scored once. No filtering is applied to the tables.

| configs | queries | hard negatives | teacher scores |
|---|---:|---:|---:|
| `hard-negatives` · `teacher-scores` | 100,000 (seeded sample, seed 1) | 9,889,255 (9,889,255 dense) | 9,989,255 |

```python
from datasets import load_dataset
negatives = load_dataset("Hyukkyu/beir-fever", "hard-negatives", split="train")
scores    = load_dataset("Hyukkyu/beir-fever", "teacher-scores", split="train")
```

## Jev judgments

The `judgments` config holds, for every training query, whether [TypeSafe](https://typesafe.ai)'s Jev (`jev-1.13.0`) judged its training positive and its mined candidates relevant: `p_yes` is Jev's P(yes) for the source's question (e.g. *does the passage answer the query?*). They locate the false negatives among the mined candidates and the mislabelled positives.

- **Requests.** One request per query: its training positive, its candidates whose teacher score is at least 0.85 × the positive's (the 24 highest) and 8 random candidates below that, shuffled under neutral ids, one yes/no question per passage (`round` 0). Queries left with fewer than 7, then 10, candidates under the cutoff got the next hardest unjudged candidates in top-up rounds (`round` 1–8). The dataset's own labelled negatives were never sent. Texts were cut to 256 (query) and 512 (passage) tokens of the `jina-embeddings-v5` small tokenizer. Grouped judgments run about 0.07 below single-pair ones, so the thresholds below apply to this table.
- **Accuracy** (blind-labelled audit of 1,946 pairs): a candidate at or above its source's cutoff is relevant 73% of the time, one below it 9%; a positive under 0.15 is mislabelled 80% of the time, except in AG News, NPR and HotpotQA, whose positives are right by construction (Jev's flags there were 0–17% right).
- **Use** (the SPARSE loader, `annotation.filter.judge`): a candidate at P(yes) ≥ the cutoff is never a negative; a positive under 0.15 is replaced by the candidate Jev scores highest if that is ≥ 0.8, else the query is dropped; a candidate at ≥ 0.9 can become an extra positive.

| config | queries | rows | candidates per query | top-up rows | candidate cutoff | positive check |
|---|---:|---:|---:|---:|---:|---|
| `judgments` | 100,000 | 2,366,979 | 22.7 | 1,320,973 | 0.58 | < 0.15 |

## Load it

```python
from datasets import load_dataset
queries = load_dataset("Hyukkyu/beir-fever", "queries", split="test")
corpus  = load_dataset("Hyukkyu/beir-fever", "corpus", split="test")
qrels   = load_dataset("Hyukkyu/beir-fever", "qrels", split="test")
```

## Cross-check against the previous layout

This repo previously held the same dataset in BEIR's raw two-repo form (`Hyukkyu/beir-fever` + `Hyukkyu/beir-fever-qrels`; that layout is kept under the git tag `legacy-layout`). Before replacing it, the build was compared against it: corpus id set **identical** (5,416,568 documents); qrels rows per split `test` 7,937 = 7,937, `train` 140,085 = 140,085, `dev` 8,079 = 8,079.

## License and attribution

The data is redistributed under the source's terms — `cc-by-nc-sa-3.0`. All credit belongs to the
original authors; see the source repository and the references in `mteb`'s task metadata.
This repository is an independent repackaging and is not affiliated with the RTEB or MTEB maintainers.

> **License discrepancy.** The upstream dataset card declares `cc-by-nc-sa-3.0` while `mteb`'s `TaskMetadata` declares `None`. This card carries the upstream value as the more conservative choice.