---
title: "South Bend — research institutions and their papers"
description: "South Bend (United States): 1 institutions with papers on Elamur Research — 1 university — and 35,879 citations of their works."
canonical: "https://research.elamur.ai/free/en/agglomerations/south-bend/"
---
[Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/agglomerations/south-bend/)

[home](https://md.research.elamur.ai/free/en/) / [institutions](https://md.research.elamur.ai/free/en/orgs/) / [agglomerations](https://md.research.elamur.ai/free/en/agglomerations/) / South Bend

# South Bend

Agglomeration [United States](https://md.research.elamur.ai/free/en/countries/united-states/) · metro population 558,370 (functional urban area, GHSL 2015)

**1** institution with a page**19** works in our library**35,879** citations of their works

## About

### Profile

South Bend is a city region in the [United States](https://md.research.elamur.ai/free/en/countries/united-states/). On this site its research is carried by one institution, the [University of Notre Dame](https://md.research.elamur.ai/free/en/universities/university-of-notre-dame/), a university located in Notre Dame. All the region's works and citations here come from that university, so South Bend's scientific profile on this site is the profile of Notre Dame.

That profile is narrow and coherent. Most of the works here belong to computer science, and above all to natural language processing and the study of large language models. A second, older line in machine learning concerns learning from imbalanced data. Together the two show a region whose visible research, on this site, centres on how machines learn from data and how far their outputs can be trusted.

### History

The earliest work here is from 2002: [SMOTE: Synthetic Minority Over-sampling Technique](https://md.research.elamur.ai/free/en/papers/chawla-2002-smote-synthetic-minority-over-sampling/), led by [Nitesh V. Chawla](https://md.research.elamur.ai/free/en/authors/nitesh-v-chawla/). It addresses a basic machine-learning problem. When one class in a dataset is much rarer than another, classifiers tend to ignore it. For a long stretch after it, this method is the region's main presence on this site.

The next visible line begins in 2018, in speech and machine translation, with [Tied Multitask Learning for Neural Speech Translation](https://md.research.elamur.ai/free/en/papers/anastasopoulos-2018-tied-multitask-learning-neural-speech/) by [Antonios Anastasopoulos](https://md.research.elamur.ai/free/en/authors/antonios-anastasopoulos/). From then on the works shift toward language technology. Work on scoring text with masked language models follows in 2020. An audit of multilingual web corpora follows in 2021.

From 2022 the focus moves to large language models. In 2022 comes a study of generation as an alternative to retrieval. In 2023 comes a study of privacy attacks on chat systems. In 2024 come two broad works: a survey of multi-agent systems built on language models, and a benchmark of trustworthiness. So the research here runs from classical machine learning to multilingual NLP and then to the behaviour, uses and risks of large models.

### Research profile

Machine learning from imbalanced data is the oldest strand. SMOTE handles a rare class by creating new synthetic examples for it. Copying the existing minority cases would only repeat them, so the method generates artificial ones instead. This gives a classifier a richer picture of the rare class. Among the works here it stands as the clearest contribution to general machine-learning method.

Speech and machine translation form the second strand. The 2018 work on tied multitask learning proposes a model shaped like a triangle. In it, the translation decoder draws on two sources at once: the encoder's representation of the audio frames, and an intermediate transcription. Recognising and translating speech are learned together rather than as separate steps. The work belongs to research on translating speech directly, including for languages with few resources.

Language-model evaluation and data quality make up a third strand. [Masked Language Model Scoring](https://md.research.elamur.ai/free/en/papers/salazar-2020-masked-language-model-scoring/) shows how to score sentences with masked language models, which were not designed to assign probabilities to text. It also introduces pseudo-perplexity as a way to evaluate such a model on a whole corpus. [Quality at a Glance: An Audit of Web-Crawled Multilingual Datasets](https://md.research.elamur.ai/free/en/papers/kreutzer-2021-quality-glance-audit-web-crawled/) examines large multilingual corpora gathered from the web. It argues that poor-quality releases can make progress on low-resource languages look greater than it really is. Both works ask how the field measures what language models and their training data actually deliver.

Large language models as knowledge sources form the fourth strand. [Generate rather than Retrieve: Large Language Models are Strong Context Generators](https://md.research.elamur.ai/free/en/papers/yu-2022-generate-rather-than-retrieve-large/) presents GenRead. Most question-answering systems first search a document collection for useful passages. GenRead drops that retriever and asks a large language model, InstructGPT, to write the supporting context itself. The work shows that context generated this way can stand in for retrieved documents on knowledge-heavy questions.

Safety, privacy and trust are the fifth strand. [Multi-step Jailbreaking Privacy Attacks on ChatGPT](https://md.research.elamur.ai/free/en/papers/li-2023-multi-step-jailbreaking-privacy-attacks/) tests whether a chat system can be led to reveal personal information. It pairs names with email addresses and phone numbers, drawn partly from the public Enron Email Dataset. [TrustLLM: Trustworthiness in Large Language Models](https://md.research.elamur.ai/free/en/papers/huang-2024-trustllm-trustworthiness-large-language-models/) responds to the absence of a common way to judge whether such models behave reliably. It offers a unified framework and benchmark for that judgement, and it rests on a very long reference list.

Agents built on language models form the sixth strand. [Large Language Model based Multi-Agents: A Survey of Progress and Challenges](https://md.research.elamur.ai/free/en/papers/guo-2024-large-language-model-based/) proposes a framework for describing any system in which several language-model agents work together. It sets out the dimensions of such a system, beginning with how the agents connect to their environment, and uses them to compare existing approaches.

### Leading scientists

Nitesh V. Chawla comes first among the people of the region's university. On this site Chawla's best-known contribution is the 2002 SMOTE method for imbalanced classification. It is the foundation of the region's machine-learning strand.

[David Chiang](https://md.research.elamur.ai/free/en/authors/david-chiang/) follows. Chiang's profile on this site gathers that researcher's own works at the university. They sit alongside the language-technology research that dominates the region's output here.

[Wenhao Yu](https://md.research.elamur.ai/free/en/authors/wenhao-yu/) is the first author of the 2022 GenRead study. That work brings large language models into open-domain question answering as generators of context rather than consumers of retrieved text.

[Toan Q. Nguyen](https://md.research.elamur.ai/free/en/authors/toan-q-nguyen/) comes next. Nguyen's works at Notre Dame appear on that researcher's own profile on this site, within the university's language-processing research.

Antonios Anastasopoulos led the 2018 work on tied multitask learning for speech translation. It is the region's clearest contribution on this site to translating spoken language with neural models.

[Meng Jiang](https://md.research.elamur.ai/free/en/authors/meng-jiang/), [Xiangliang Zhang](https://md.research.elamur.ai/free/en/authors/xiangliang-zhang/) and [Kenton Murray](https://md.research.elamur.ai/free/en/authors/kenton-murray/) complete the group of the university's first people here. Each has a profile on this site gathering their works. Their presence fits the strong turn of the region's recent output toward large language models, their evaluation and their trustworthiness.

### Landmark works

Several works account for most of the region's influence on this site. Among them, the Kreutzer audit of multilingual web-crawled datasets and the Salazar study of masked language model scoring are the most cited by other papers here. Both are about measurement: one about the quality of training data, the other about how to evaluate language models.

SMOTE is close behind them and remains the region's most visible classical machine-learning result. Its idea of creating synthetic minority examples, instead of duplicating existing ones, has made it a standard reference point for imbalanced learning among the works here.

The 2024 survey of language-model-based multi-agent systems is also widely cited here, and so is the 2022 GenRead study. Together they mark the region's engagement with the newest uses of large models. One work maps systems in which several agents cooperate. The other rethinks the retrieval step in question answering.

TrustLLM and the multi-step jailbreaking study complete the set of most-cited works, on the side of risk. One builds a broad framework for judging the reliability of large models. The other shows a concrete route by which private details might be extracted from them. The 2018 speech-translation paper is a smaller but distinct landmark, rooted in the earlier speech and translation strand.

### Labs and units

On this site the region's research is organised under a single institution, the University of Notre Dame. No separate departments, laboratories or member institutions are shown for the city region. Its scientists and works are grouped at the level of the university.

### Collaborations

The region has only one institution with works here, a university, and no companies. Links between local industry and academia therefore cannot be traced on this site.

Several of the most cited works have first authors who are not among the university's first people here, including Kreutzer, Salazar, Li, Guo and Huang. Notre Dame's contribution to them therefore sits within wider author teams. This is especially clear for large efforts such as TrustLLM and the multilingual dataset audit. Within the United States, South Bend's role in these joint works is that of a university partner in collaborative language-technology research.

### Timeline

In 2002 the region's visible research begins with SMOTE. That method for learning from imbalanced classes stands alone on this site for many years.

In 2018 the work turns to language, with the tied multitask model for neural speech translation. It opens the speech and translation line.

In 2020 and 2021 come evaluation and data. The 2020 study shows how to score text with masked language models. The 2021 audit questions the quality of web-crawled multilingual corpora.

From 2022 to 2024 large language models take centre stage. GenRead appears in 2022 and the privacy-attack study on ChatGPT in 2023. In 2024 come the survey of multi-agent systems and the TrustLLM framework. On this site, these years show South Bend's research concentrated on what large language models can do and how far they can be trusted.

## Institutions (1)

1 university.

### 🎓 Universities (1)

1. **[University of Notre Dame](https://md.research.elamur.ai/free/en/universities/university-of-notre-dame/)**University · Notre Dame, [United States](https://md.research.elamur.ai/free/en/countries/united-states/)*19 works · 35,879 citations*

## Works in our library (19)

Works of its institutions, the most cited first.

1. [**SMOTE: Synthetic Minority Over-sampling Technique**Nitesh V. Chawla, Kevin W. Bowyer, Lawrence Hall, et al.*2002 · 32,661 citations*](https://md.research.elamur.ai/free/en/papers/chawla-2002-smote-synthetic-minority-over-sampling/)
2. [**Quality at a Glance: An Audit of Web-Crawled Multilingual Datasets**Julia Kreutzer, Isaac Caswell, Lisa Wang, et al.*2021 · 447 citations*](https://md.research.elamur.ai/free/en/papers/kreutzer-2021-quality-glance-audit-web-crawled/)
3. [**Masked Language Model Scoring**Julian Salazar, Davis Liang, Toan Q. Nguyen, et al.*2020 · 444 citations*](https://md.research.elamur.ai/free/en/papers/salazar-2020-masked-language-model-scoring/)
4. [**Multi-step Jailbreaking Privacy Attacks on ChatGPT**Haoran Li, Dadi Guo, Wei Fan, et al.*2023 · 298 citations*](https://md.research.elamur.ai/free/en/papers/li-2023-multi-step-jailbreaking-privacy-attacks/)
5. [**Large Language Model based Multi-Agents: A Survey of Progress and Challenges**Taicheng Guo, Xiuying Chen, Yaqi Wang, et al.*2024 · 282 citations*](https://md.research.elamur.ai/free/en/papers/guo-2024-large-language-model-based/)
6. [**Generate rather than Retrieve: Large Language Models are Strong Context Generators**Wenhao Yu, Dan Iter, Shuohang Wang, et al.*2022 · 280 citations*](https://md.research.elamur.ai/free/en/papers/yu-2022-generate-rather-than-retrieve-large/)
7. [**Tied Multitask Learning for Neural Speech Translation**Antonios Anastasopoulos, David Chiang*2018 · 250 citations*](https://md.research.elamur.ai/free/en/papers/anastasopoulos-2018-tied-multitask-learning-neural-speech/)
8. [**TrustLLM: Trustworthiness in Large Language Models**Yue Huang, Lichao Sun, Haoran Wang, et al.*2024 · 226 citations*](https://md.research.elamur.ai/free/en/papers/huang-2024-trustllm-trustworthiness-large-language-models/)
9. [**Improved Neural Machine Translation with a Syntax-Aware Encoder and Decoder**Huadong Chen, Shujian Huang, David Chiang, et al.*2017 · 192 citations*](https://md.research.elamur.ai/free/en/papers/chen-2017-improved-neural-machine-translation-syntax/)
10. [**Auto-Debias: Debiasing Masked Language Models with Automated Biased Prompts**Yue Guo, Yi Yang, Ahmed Abbasi*2022 · 191 citations*](https://md.research.elamur.ai/free/en/papers/guo-2022-auto-debias-debiasing-masked-language/)
11. [**BigCodeBench: Benchmarking Code Generation with Diverse Function Calls and Complex Instructions**Terry Yue Zhuo, Minh Chien Vu, Jenny Chim, et al.*2024 · 163 citations*](https://md.research.elamur.ai/free/en/papers/zhuo-2024-bigcodebench-benchmarking-code-generation-diverse/)
12. [**Justice or Prejudice? Quantifying Biases in LLM-as-a-Judge**Jiayi Ye, Yanbo Wang, Yue Huang, et al.*2024 · 142 citations*](https://md.research.elamur.ai/free/en/papers/ye-2024-justice-prejudice-quantifying-biases-llm/)
13. [**Fantastic Questions and Where to Find Them: FairytaleQA – An Authentic Dataset for Narrative Comprehension**Ying Xu, Dakuo Wang, Mo Yu, et al.*2022 · 136 citations*](https://md.research.elamur.ai/free/en/papers/xu-2022-fantastic-questions-where-find-them/)
14. [**It is AI's Turn to Ask Humans a Question: Question-Answer Pair Generation for Children's Story Books**Bingsheng Yao, Dakuo Wang, Tongshuang Wu, et al.*2022 · 64 citations*](https://md.research.elamur.ai/free/en/papers/yao-2022-it-is-ai-s-turn/)
15. [**LegalAgentBench: Evaluating LLM Agents in Legal Domain**Haitao Li, Junjie Chen, Jingli Yang, et al.*2025 · 60 citations*](https://md.research.elamur.ai/free/en/papers/li-2025-legalagentbench-evaluating-llm-agents-legal/)
16. [**IfQA: A Dataset for Open-domain Question Answering under Counterfactual Presuppositions**Wenhao Yu, Meng Jiang, Peter Clark, et al.*2023 · 26 citations*](https://md.research.elamur.ai/free/en/papers/yu-2023-ifqa-dataset-open-domain-question/)
17. [**ArMATH: a Dataset for Solving Arabic Math Word Problems**Reem Alghamdi, Zhenwen Liang, Xiangliang Zhang*2022 · 17 citations*](https://md.research.elamur.ai/free/en/papers/alghamdi-2022-armath-dataset-solving-arabic-math/)
18. [**MiqraBERT: Regression-Based Sentence-BERT Finetuning for Biblical Hebrew Parallel Detection**David M. Smiley*2026*](https://md.research.elamur.ai/free/en/papers/smiley-2026-miqrabert-regression-based-sentence-bert/)
19. [**T'OMIM: A Morphologically Annotated Dataset of Parallel Passages in the Hebrew Bible**David M. Smiley*2026*](https://md.research.elamur.ai/free/en/papers/smiley-2026-tomim-morphologically-annotated-dataset-parallel/)

## Works by year

### 2020s

[**2026** 2 works](https://md.research.elamur.ai/free/en/agglomerations/south-bend/year/2026/) · [**2025** 1 work](https://md.research.elamur.ai/free/en/papers/li-2025-legalagentbench-evaluating-llm-agents-legal/) · [**2024** 4 works](https://md.research.elamur.ai/free/en/agglomerations/south-bend/year/2024/) · [**2023** 2 works](https://md.research.elamur.ai/free/en/agglomerations/south-bend/year/2023/) · [**2022** 5 works](https://md.research.elamur.ai/free/en/agglomerations/south-bend/year/2022/) · [**2021** 1 work](https://md.research.elamur.ai/free/en/papers/kreutzer-2021-quality-glance-audit-web-crawled/) · [**2020** 1 work](https://md.research.elamur.ai/free/en/papers/salazar-2020-masked-language-model-scoring/)

### 2010s

[**2018** 1 work](https://md.research.elamur.ai/free/en/papers/anastasopoulos-2018-tied-multitask-learning-neural-speech/) · [**2017** 1 work](https://md.research.elamur.ai/free/en/papers/chen-2017-improved-neural-machine-translation-syntax/)

### 2000s

[**2002** 1 work](https://md.research.elamur.ai/free/en/papers/chawla-2002-smote-synthetic-minority-over-sampling/)

## Concepts, shelves and topics

Of its works in our library: the number of works on each and its rating among them.

📚 Shelves 8[AI & machine learning · 2 works](https://md.research.elamur.ai/free/en/shelves/ai/) · [Natural language processing · 7 works](https://md.research.elamur.ai/free/en/shelves/nlp/) · [Other · 5 works](https://md.research.elamur.ai/free/en/shelves/other/) · [Machine translation · 1 work](https://md.research.elamur.ai/free/en/shelves/machine-translation/) · [Linguistics & languages · 5 works](https://md.research.elamur.ai/free/en/shelves/linguistics/) · [English · 2 works](https://md.research.elamur.ai/free/en/shelves/en/) · [Law · 1 work](https://md.research.elamur.ai/free/en/shelves/law/) · [Ancient Hebrew · 2 works](https://md.research.elamur.ai/free/en/shelves/hbo/)

🧭 Topics 15[Language models & pretraining · 5 works](https://md.research.elamur.ai/free/en/topics/language-models/) · [Bias, fairness & safety · 3 works](https://md.research.elamur.ai/free/en/topics/bias-fairness-safety/) · [Multilingual & cross-lingual NLP · 1 work](https://md.research.elamur.ai/free/en/topics/multilingual-nlp/) · [Speech & spoken language · 1 work](https://md.research.elamur.ai/free/en/topics/speech/) · [Code & programming languages · 1 work](https://md.research.elamur.ai/free/en/topics/code-and-programming/) · [Information retrieval & search · 1 work](https://md.research.elamur.ai/free/en/topics/information-retrieval/) · [Legal AI · 1 work](https://md.research.elamur.ai/free/en/topics/legal-ai/) · [Machine translation · 1 work](https://md.research.elamur.ai/free/en/topics/machine-translation/) · [Syntax & grammar formalisms · 1 work](https://md.research.elamur.ai/free/en/topics/syntax-grammar/) · [Implicature · 1 work](https://md.research.elamur.ai/free/en/topics/implicature/) · [Question answering · 1 work](https://md.research.elamur.ai/free/en/topics/question-answering/) · [Semitic languages · 3 works](https://md.research.elamur.ai/free/en/topics/semitic-languages/) · [Reasoning · 1 work](https://md.research.elamur.ai/free/en/topics/reasoning/) · [Hebrew & Aramaic · 2 works](https://md.research.elamur.ai/free/en/topics/hebrew-aramaic/) · [Morphology · 1 work](https://md.research.elamur.ai/free/en/topics/morphology/)

Agglomerations: the GHSL Urban Centre Database 2025; populations: the GHSL functional urban areas 2015 (urban centre and commuting zone), else the urban centre, as each page says (European Commission, Joint Research Centre, CC BY 4.0). An institution belongs to the urban centre its coordinates (ROR, OpenAlex) fall in, or one within 5 km. Land: Natural Earth.

- [← back to](https://md.research.elamur.ai/free/en/countries/united-states/) — United States

Updated 23 October 2026