# Language Research

## Research Scientists & Research Interns @ Language Research, NAVER AI Lab

### Join Our Team!

[About Us](#about-us)

[About the Research Scientists](#about-the-research-scientists)

[About the Research Interns](#about-the-research-interns)

### **About Us**

**Language Research Team** at [NAVER AI Lab](https://naver-career.gitbook.io/en/teams/clova-cic/ai-lab) is dedicated to understanding humanity and society, and advancing human-like but also trustworthy and safe language models and Artificial Intelligence. As a team operating in both academic and industrial environments, we strive to tackle problems that are both fundamental and relevant to the real world.

Our current **Research Mission and Interests** are centered around building trustworthy and safe Large Language Models (LLMs), with a focus on:

* Datasets, Benchmarks, and Evaluation Metrics for LLMs
* LLM Security: Attacks, Defenses & Detections
* Safety Alignment, Learning, and Inference Algorithms
* LLM Agents, (Multi-)Agent Interactions, Decision-making, and Autonomous Agents

Check out our latest papers\*([selected papers](#selected-papers), [all](https://naver-career.gitbook.io/en/publications/all)).\*

***

### About the Research Scientists

#### About the Role

We are looking for Research Scientists to join our team to research and development of **safe and trustworthy Language Models and AI**. The Research Scientists are encouraged to lead and/or support research projects collaboratively within the team and across the research field, other teams, and external organizations.

Specifically, the research topics include following but not limited to:

* Red-teaming, Adversarial Attack, Security Attack
* Watermarking
* Training Data/Privacy Probing & Leakage
* Model/Data/Task Contamination
* Robustness
* Safety Alignment
* Model Unlearning
* AI Explainability & Interpretability
* Causality
* Societal Impact by LLM Applications

#### Key Responsibilities

* Undertake pioneering research by formulating challenging research questions and devising problem-solving methods.
* Lead a wide range of research activities including but not limited to the ideation and development of safe and trustworthy AI systems, and authoring research papers.
* Communicate research progress and findings clearly and effectively.
* Actively collaborate with other researchers.
* Report and present the research findings and developments at top-tier academic venues.

#### Requirements

* Holds a PhD degree or equivalent (or expected to receive within 6 months) in Computer Science (CS), Electrical Engineering (EE), Mathematics, or other relevant fields.
* An academic publication record at top-tier conferences in Natural Language Processing (e.g., \*ACL), Machine Learning (e.g., NeurIPS, ICLR), and others (e.g., FAccT).
* Experience in research collaborations and academic writing in related fields.
  * **(Preferred)** Global research/industrial collaboration experiences.
* Excellent analytical and problem-solving skills.
* Strong communication skills, openness to constructive discussion, and receptiveness to feedback.

#### How to apply

* We are planning to reopen the application. Thank you for your understanding.

***

### [About the Research Interns](https://hwaranlee.notion.site/Research-Intern-Language-Research-NAVER-AI-Lab-6d518d575af04963866cf0ff309b798e?pvs=4)

#### About the Internship

Our team is offering research intern positions for <mark style="color:blue;">**2024 Fall and 2025 Winter**</mark>. As an intern, you'll be actively involved in developing and conducting research on **trustworthy and safe large language models**.

**Before starting your internship, we will discuss closely to refine and develop your research plan.** This process ensures that your proposal aligns with our mutual research interests. We strongly support your initiative to lead your main project while also engaging in other research projects. This approach offers a balanced experience in both research leadership and collaboration.

A key goal of this internship is to produce academic papers suitable for submission to top-tier conferences or journals. Additionally, we anticipate that the outcomes of the project will make meaningful contributions to real-world applications.

* This is a full-time, in-person role at [**NAVER 1784**](https://www.navercorp.com/en/naver/1784) (Seongnam-si, Gyeonggi-do, South Korea)
  * The office could be changed to NAVER Green Factory or the other near building.
* This internship offers a flexible starting date.

#### Key Responsibilities

* Undertake pioneering research by formulating challenging research questions and devising problem-solving methods. This includes implementing and evaluating models, as well as authoring research papers.
* Communicate research progress and findings clearly and effectively.
* Demonstrate proactivity and the ability to successfully complete projects.

#### Requirements

* Pursuing a PhD or equivalent in Computer Science (CS), Electrical Engineering (EE), Mathematics, or other relevant fields.
* At least one paper authored as the first author in AI/ML-related conferences.
  * **(Preferred)** A strong academic publication record at top-tier conferences in Natural Language Processing (e.g., \*CL), Machine Learning (e.g., NeurIPS, ICLR), and others.
* Experience in research collaborations and academic writing in related fields.
* Excellent analytical and problem-solving skills.
* Strong communication skills, openness to constructive discussion, and receptiveness to feedback.\`

#### How to apply&#x20;

* We are planning to reopen the application. Thank you for your understanding.

We look forward to your application and the possibility of you joining our team. If you have any question, please contact us! 🤗

***

### Selected Papers

* [**MAQA: Evaluating Uncertainty Quantification in LLMs Regarding Data Uncertainty,**](https://naver-career.gitbook.io/en/positions/ai-ml/broken-reference) Yongjin Yang, Haneul Yoo, Hwaran Lee, **Arxiv**, *2024* <mark style="color:red;">`dataset & benchmark`</mark><mark style="color:green;">`uncertainty`</mark>
* [**CSRT: Evaluation and Analysis of LLMs using Code-Switching Red-Teaming Dataset**](https://arxiv.org/abs/2406.15481)**,** Haneul Yoo, Yongjin Yang, Hwaran Lee, **Arxiv**, *2024* <mark style="color:red;">`dataset & benchmark`</mark>` `<mark style="color:green;">`llm-security`</mark>
* [**AdvisorQA: Towards Helpful and Harmless Advice-seeking Question Answering with Collective Intelligence**](https://arxiv.org/abs/2404.11826)\
  **,** Minbeom Kim, Hwanhee Lee, Joonsuk Park, Hwaran Lee, Kyomin Jung, **Arxiv**, *2024* <mark style="color:purple;">`alignment`</mark>` `<mark style="color:red;">`dataset & benchmark`</mark>
* [**Who Wrote this Code? Watermarking for Code Generation**](https://arxiv.org/abs/2305.15060)**,** Taehyun Lee, Seokhee Hong, Jaewoo Ahn, Ilgee Hong, Hwaran Lee, Sangdoo Yun, Jamin Shin, Gunhee Kim, **ACL**, *2024* <mark style="color:green;">`llm-security`</mark>
* [**Calibrating Large Language Models Using Their Generations Only**](https://arxiv.org/abs/2403.05973)**,** Dennis Thomas Ulmer, Martin Gubri, Hwaran Lee, Sangdoo Yun, Seong Joon Oh, **ACL**, *2024* <mark style="color:green;">`uncertainty`</mark>
* [**TRAP: Targeted Random Adversarial Prompt Honeypot for Black-Box Identification**](https://arxiv.org/abs/2402.12991)**,** Martin Gubri, Dennis Thomas Ulmer, Hwaran Lee, Sangdoo Yun, Seong Joon Oh, **ACL Findings**, *2024* <mark style="color:green;">`llm-security`</mark>
* [**KorNAT: LLM Alignment Benchmark for Korean Social Values and Common Knowledge**](https://arxiv.org/abs/2402.13605)**,** Jiyoung Lee, Minwoo Kim, Seungho Kim, Junghwan Kim, Seunghyun Won, Hwaran Lee, Edward Choi, **ACL Findings**, *2024* <mark style="color:red;">`dataset & benchmark`</mark>
* [**TimeChara: Evaluating Point-in-Time Character Hallucination of Role-Playing Large Language Models**](https://arxiv.org/abs/2405.18027)**,** Jaewoo Ahn, Taehyun Lee, Junyoung Lim, Jin-Hwa Kim, Sangdoo Yun, Hwaran Lee, Gunhee Kim, **ACL Findings**, *2024* <mark style="color:red;">`dataset & benchmark`</mark>
* [**LifeTox: Unveiling Implicit Toxicity in Life Advice**](https://arxiv.org/abs/2311.09585), M Kim, J Koo, H Lee, J Park, H Lee, K Jung, **NAACL (Shrot)** <mark style="color:red;">`dataset & benchmark`</mark>
* [**Prometheus: Inducing Fine-grained Evaluation Capability in Language Models**](https://arxiv.org/abs/2310.08491), S Kim, J Shin, Y Cho, J Jang, S Longpre, H Lee, S Yun, S Shin, S Kim, J Throne, M Seo, **ICLR 2024** <mark style="color:red;">`dataset`</mark> <mark style="color:blue;">`evaluation`</mark>
* [**EvalLM: Interactive Evaluation of Large Language Model Prompts on User-Defined Criteria**](https://arxiv.org/abs/2309.13633), TS Kim, Y Lee, J Shin, YH Kim, J Kim, **arXiv preprint arXiv:2309.13633** <mark style="color:blue;">`evaluation`</mark>
* [**KoBBQ: Korean Bias Benchmark for Question Answering**](https://arxiv.org/abs/2307.16778), J Jin, J Kim, N Lee, H Yoo, A Oh, H Lee, **TACL**  <mark style="color:red;">`dataset & benchmark`</mark>
* [**Revealing User Familiarity Bias in Task-Oriented Dialogue via Interactive Evaluation**](https://arxiv.org/abs/2305.13857), T Kim, J Shin, YH Kim, S Bae, S Kim, **arXiv preprint arXiv:2305.13857** <mark style="color:blue;">`evaluation`</mark>
* [**The CoT Collection: Improving Zero-shot and Few-shot Learning of Language Models via Chain-of-Thought Fine-Tuning**](https://arxiv.org/abs/2305.14045), S Kim, SJ Joo, D Kim, J Jang, S Ye, J Shin, M Seo, **EMNLP 2023** <mark style="color:red;">`dataset`</mark>
* [**Aligning Large Language Models through Synthetic Feedback**](https://arxiv.org/abs/2305.13735), S Kim, S Bae, J Shin, S Kang, D Kwak, KM Yoo, M Seo, **EMNLP 2023** <mark style="color:purple;">`alignment`</mark>
* [**ProPILE: Probing Privacy Leakage in Large Language Models**](https://arxiv.org/abs/2307.01881), S Kim, S Yun, H Lee, M Gubri, S Yoon, SJ Oh, **NeurIPS 2023&#x20;*****(spotlight)*** <mark style="color:green;">`llm-security`</mark>
* [**Who Wrote this Code? Watermarking for Code Generation**](https://arxiv.org/abs/2305.15060), T Lee, S Hong, J Ahn, I Hong, H Lee, S Yun, J Shin, G Kim, **arXiv preprint arXiv:305.15060** <mark style="color:green;">`llm-security`</mark>
* [**KoSBi: A Dataset for Mitigating Social Bias Risks Towards Safer Large Language Model Application**](https://arxiv.org/abs/2305.17701), H Lee, S Hong, J Park, T Kim, G Kim, JW Ha, **ACL 2023&#x20;*****(industry track)*** <mark style="color:red;">`dataset & benchmark`</mark>
* [**SQuARe: A Large-Scale Dataset of Sensitive Questions and Acceptable Responses Created Through Human-Machine Collaboration**](https://arxiv.org/abs/2305.17696), H Lee, S Hong, J Park, T Kim, M Cha, Y Choi, BP Kim, G Kim, EJ Lee, Y Lim, A Oh, S Park, JW Ha, **ACL 2023&#x20;*****(best paper nominated)*** <mark style="color:red;">`dataset & benchmark`</mark>
* [**Query-Efficient Black-Box Red Teaming via Bayesian Optimization**](https://arxiv.org/abs/2305.17444), D Lee, JY Lee, JW Ha, JH Kim, SW Lee, H Lee, HO Song, **ACL 2023** <mark style="color:green;">`llm-security`</mark>
* [**Critic-Guided Decoding for Controlled Text Generation**](https://arxiv.org/abs/2212.10938), M Kim, H Lee, KM Yoo, J Park, H Lee, K Jung, **ACL 2023 (Findings)** <mark style="color:purple;">`learning & inference`</mark>
* [**ClaimDiff: Comparing and Contrasting Claims on Contentious Issues**](https://aclanthology.org/2023.findings-acl.289/), M Ko, I Seong, H Lee, J Park, M Chang, M Seo, **ACL 2023(Findings)** <mark style="color:red;">`dataset & benchmark`</mark>