Datasets:
candidate_id int64 146k 283M ⌀ | title large_stringlengths 11 300 ⌀ | abstract large_stringlengths 50 10k ⌀ | field_of_study large_stringclasses 25
values | citation_count int64 1 4 ⌀ | body large_stringlengths 8k 79.9k ⌀ |
|---|---|---|---|---|---|
213,808,097 | A Mineralisation Age for the Sediment-Hosted Blackbush Uranium Prospect, North-Eastern Eyre Peninsula, South Australia | The Blackbush uranium prospect (~12,580 tonnes U at 85 ppm cut-off) is located on the Eyre Peninsula of South Australia. Blackbush was discovered in 2007 and is currently the single example of sediment-hosted uranium mineralisation investigated in any detail in the Gawler Craton. Uranium is hosted within Eocene sandsto... | Geology | 2 |
Introduction
South Australia has long been recognised as being exceptionally well endowed with uranium resources. All currently economically relevant deposits are associated with high-U Mesoproterozoic felsic rocks [1]. In the Gawler Craton, iron-oxide copper-gold (IOCG) deposits are genetically related to the emplac... |
272,572,181 | Trust Us—We Are the (COVID-19 Misinformation) Experts: A Critical Scoping Review of Expert Meanings of “Misinformation” in the Covid Era | Since the WHO declared COVID-19 a pandemic, prominent social actors and institutions have warned about the threat of misinformation, calling for policy action to address it. However, neither the premises underlying expert claims nor the standards to separate truth from falsehood have been appraised. We conducted a scop... | Unknown | 3 |
Introduction
Since the World Health Organization (WHO) announced to the world, in March 2020, that a new disease, COVID-19, had reached the status of pandemic [1], dominant institutions, groups, and individuals (hereafter "the establishment") have voiced their concern about the existential threat posed by false or mi... |
55,260,885 | BRAZILIAN INITIAL PUBLIC OFFERINGS, UNDERWRITERS, AND PREMIUM CORPORATE GOVERNANCE SEGMENTS LISTING | We examined 89 offers in the most recent Brazilian IPO wave between 2004 and 2007, all listed in premium segments of the exchange that demand better corporate governance practices. Two non-US underwriters dominated the market, often acting as co-leaders and rarely as second-tier underwriters. Twenty-eight percent of is... | Business | 1 |
The uniqueness of this Brazilian IPO wave motivated the examining of the quantity of lead underwriters, their compensation and relationships in the syndicates. We relate them to IPO characteristics, such as initial market-adjusted returns, ex-ante uncertainty, and offer size. Sixty-eight percent of IPOs had more than ... |
214,584,636 | Mapping the reach of a rural Transitions Nurse Program for veterans with geographic information systems | Rural Veterans who receive inpatient care at a Veterans Health Administration (VA) tertiary facility can face significant barriers to a safe transition home. The VA rural Transitions Nurse Program (TNP) is a national, intensive care coordination quality improvement program for rural Veterans. To communicate the reach o... | Geography | 3 |
Background
Nearly 2.7 million Veterans live in rural communities and use the Veterans Health Administration (VA) for healthcare [1]. Access to care for this population is a challenge due to significant geographic distances between VA tertiary hospitals, located in urban settings, and VA clinics and hospitals, commonl... |
28,031,850 | A Preliminary Study for Building an Arabic Corpus of Pair Questions-texts from the Web: AQA-WebCorp | With the development of electronic media and the heterogeneity of Arabic data on the Web, the idea of building a clean corpus for certain applications of natural language processing, including machine translation, information retrieval, question answer, become more and more pressing. In this manuscript, we seek to crea... | Computer Science | 3 |
INTRODUCTION
The lack and / or the absence of corpus in Arabic has been a problem for the implementation of natural language processing. This also has a special interest in the track of the question answering.
The corpus (singular form of corpora 1 ) construction is a task that both essential and delicate. It is co... |
271,291,482 | Gender bias in child custody judgments: Evidence from Chinese family court | "Based on a quantitative analysis of a novel dataset comprising 10,093 publicly available judgments (...TRUNCATED) | Law | 3 | "\nIntroduction\n\nThe rising divorce rate has resulted in an unprecedented surge in custody dispute(...TRUNCATED) |
118,893,245 | "Generalized Classical Dynamical Yang-Baxter Equations and Moduli Spaces of Flat Connections on Surf(...TRUNCATED) | "In this paper, we explain how generalized dynamical r-matrices can be obtained by (quasi-)Poisson r(...TRUNCATED) | Mathematics | 4 | "\nIntroduction\n\nThe classical Yang-Baxter equation (CYBE) plays a key role in the theory of integ(...TRUNCATED) |
256,107,804 | "The impact of Emotion-focused training for emotion couching delivered as mobile app on self-compass(...TRUNCATED) | "Introduction Being self-compassionate is considered a beneficial emotion regulation strategy. There(...TRUNCATED) | Psychology | 2 | "\nIntroduction\n\nAccording to Blatt and Zuroff (1992), self-criticism is constant, harsh self-scru(...TRUNCATED) |
252,550,632 | People’s Perception of Well-Being during the COVID-19 Pandemic: A Case Study in Japan | "This study aims to examine people’s perception of well-being during the COVID-19 pandemic in Japa(...TRUNCATED) | Economics | 4 | "\nIntroduction\n\nCoronavirus disease (COVID- 19), an infectious disease caused by the SARS-CoV-2 v(...TRUNCATED) |
235,540,326 | "Effects of Information Seeking Behaviour on Accessibility of Library Services by Students With Visu(...TRUNCATED) | "The major aim of the study was to investigate the effects of information seeking behaviour on acces(...TRUNCATED) | Education | 3 | "\nIntroduction\n\nLibraries and data centers around the world have created particular information s(...TRUNCATED) |
Body-Fact Retrieval (BFR) diagnostic
A science-specific long-context retrieval probe released with SciEmbed (Citation-Context Supervision for Scientific Document Embeddings, Findings of EMNLP 2026). The task: given a sentence drawn from the body of a scientific paper that does not appear in its abstract, retrieve the source paper from a pool of full-text documents. Short-context encoders that only see the title+abstract are blind to the content that must be matched; long-context encoders that read the body can recover it, so the probe isolates whether a model's context window is actually exploited on scientific text.
Contents
candidates.parquet— 9,749 candidate papers (candidate_id,title,abstract,body,field_of_study,citation_count).queries.parquet— 1,000 body-sentence queries (query_id,query_text,gold_candidate_id).
Construction
- Queries are single sentences from the middle third of each paper (skipping intro/conclusion), 80–400 characters, with at most 0.2 trigram overlap against the paper's abstract so the abstract alone is not a shortcut.
- Leakage control. Every candidate paper has a citation count in [1, 4], strictly below the citation_count ≥ 5 threshold used to build the SciEmbed Stage-2 (Signal A/B) training pool. No candidate was a training anchor, positive, or hard negative.
- Coverage. Roughly 400 papers per field across ~24 S2AG disciplines (Computer Science, Mathematics, Physics, Biology, Medicine, Law, History, Philosophy, …). Bodies span ~11k–59k characters (5th–95th percentile).
Usage
Encode each query and each candidate (title + abstract + body, up to the model's context length), then retrieve by cosine similarity and report Recall@k. See the SciEmbed repository for the full evaluation harness and the robustness variants (same-field pool, Qwen3-paraphrased queries).
Citation
@inproceedings{wilinski2026sciembed,
title={SciEmbed: Citation-Context Supervision for Scientific Document Embeddings},
author={Wilinski, Jonas and F{\"a}rber, Michael},
booktitle={Findings of the Association for Computational Linguistics: EMNLP 2026},
year={2026}
}
- Downloads last month
- -