Independent Research Projects

Overview: This project investigates methods for detecting, evaluating, and reducing hallucinations in LLMs. Students will examine factual errors, unsupported responses, and reliability across language and NLP tasks, and will evaluate practical approaches for improving model trustworthiness. The project may involve prompt-based methods, retrieval-augmented generation, automated evaluation, and comparative experiments with open and proprietary LLMs.

Project Code: IRP-01

Research Areas: Large Language Models  .  Natural Language Processing  .  Trustworthy AI  .  Hallucination Detection

Student Level: Undergraduate or Graduate

Typical Duration: 4-6 months

Minimum Qualifications: Programming proficiency, preferably in Python; knowledge of artificial intelligence, machine learning, and natural language processing; ability to read technical research papers; and sufficient availability to contribute consistently throughout the project period.

Preferred Qualifications: Prior experience with LLMs, NLP libraries, machine learning frameworks, data analysis, or research-oriented programming.

Expected Outcome: Research paper with the potential for conference or journal submission.

Status: Open to interested students

Application Process: Interested students should submit an application using Project Code IRP-01. Applications are reviewed on a rolling basis based on relevant background, technical preparation, research interests, availability, and overall fit with the project. Selected applicants will be contacted after the review process. 

Application Materials: CV/Resume and a brief Statement of Interest describing relevant skills, prior experience, motivation for the project, and expected contribution.

How to Apply: Use the "Apply for this Project" button below and submit the required materials using Project Code IRP-01.

Apply for this Project

Overview: This project investigates methods for improving the accuracy, grounding, and reliability of language models by integrating external knowledge and retrieval mechanisms. The research may explore retrieval-augmented generation (RAG), evidence selection, knowledge grounding, citation and attribution, retrieval quality, and the interaction between retrieved information and model-generated responses. Depending on the student's background and interests, the project may focus on methodological development, empirical evaluation, benchmark construction, or domain- and language-specific applications.

Project Code: IRP-02

Research Areas: Retrieval Augmented Generation  .  Knowledge Grounding  .  Large Language Models  .  Natural Language Processing

Student Level: Undergraduate or Graduate

Typical Duration: 4-6 months

Minimum Qualifications: Programming experience in Python; understanding of machine learning, NLP, or information retrieval; ability to read and summarize research papers; ability to work independently and meet project milestones.

Preferred Qualifications: Prior experience with LLMs, NLP libraries, or retrieval systems; familiarity with embeddings, vector search, or RAG pipelines; Experience with PyTorch, Hugging Face, or related tools; prior research or project experience in AI, NLP, or information retrieval.

Expected Outcome: Research report and, when results warrant, preparation of a conference or journal manuscript.

Status: Open to interested students

Application Process: Interested students should submit an application using Project Code IRP-02. Applications are reviewed on a rolling basis based on relevant background, technical preparation, research interests, availability, and overall fit with the project. Selected applicants will be contacted after the review process. 

Application Materials: CV/Resume and a brief Statement of Interest describing relevant skills, prior experience, motivation for the project, and expected contribution.

How to Apply: Use the "Apply for this Project" button below and submit the required materials using Project Code IRP-02.

Apply for this Project

Overview: This project investigates methods for making large language models more computationally efficient, accessible, and practical under limited resource constraints. The research may explore model compression, quantization, parameter-efficient adaptation, efficient inference and decoding, latency and memory optimization, cost-aware language modeling, or the trade-offs among model quality, computational resources, and deployment requirements. Depending on the researcher's background and interests, the project may focus on methodological development, empirical evaluation, benchmark construction, or resource-constrained applications of language models.

Project Code: IRP-03

Research Areas: Retrieval Augmented Generation  .  Resource-Aware AI  .  Model Compression  .  Efficient Inference  .  Efficient Large Language Models  .  Natural Language Processing

Student Level: Undergraduate or Graduate

Typical Duration: 4-6 months

Minimum Qualifications: Programming experience in Python; understanding of machine learning or NLP; ability to read and summarize research papers; ability to conduct systematic experiments and work independently toward project milestones.

Preferred Qualifications: Prior experience with LLMs, PyTorch, Hugging Face, or related NLP frameworks; familiarity with model inference; fine-tuning, quantization, or parameter-efficient methods; experience with experimental evaluation, GPU-based computing, or performance analysis.

Expected Outcome: Research report and, when results warrant, preparation of a conference or journal manuscript.

Status: Open to interested students

Application Process: Interested students should submit an application using Project Code IRP-03. Applications are reviewed on a rolling basis based on relevant background, technical preparation, research interests, availability, and overall fit with the project. Selected applicants will be contacted after the review process. 

Application Materials: CV/Resume and a brief Statement of Interest describing relevant skills, prior experience, motivation for the project, and expected contribution.

How to Apply: Use the "Apply for this Project" button below and submit the required materials using Project Code IRP-03.

Apply for this Project

Overview: This project investigates methods for developing language technologies that perform effectively across diverse languages, particularly low-resource and underrepresented languages. The research may explore multilingual and cross-lingual language models, low-resource natural language processing, machine translation, language adaptation and transfer, multilingual representation learning, terminology and meaning preservation, or methods for reducing performance disparities across languages. Depending on the researcher's background and interests, the project may focus on methodological development, empirical evaluation, benchmark or dataset construction, or applications involving specific languages, domains, and multilingual settings.

Project Code: IRP-04

Research Areas: Low-Resource Natural Language Processing  .  Cross-Lingual Learning  .  Machine Translation  .  Language Technologies  .  Large Language Models  .  Multilingual Natural Language Processing

Student Level: Undergraduate or Graduate

Typical Duration: 4-6 months

Minimum Qualifications: Programming experience in Python; understanding of machine learning or NLP; ability to read and summarize research papers; interest in multilingual or language-focused AI; ability to work independently and meet project milestones.

Preferred Qualifications: Prior experience with NLP, LLMs, machine translation, multilingual models, or language datasets; familiarity with PyTorch, Hugging Face, or related tools; experience with data preparation, model evaluation, or research-oriented programming; knowledge of more than one language is helpful but not required.

Expected Outcome: Research report and, when results warrant, preparation of a conference or journal manuscript.

Status: Open to interested students

Application Process: Interested students should submit an application using Project Code IRP-04. Applications are reviewed on a rolling basis based on relevant background, technical preparation, research interests, availability, and overall fit with the project. Selected applicants will be contacted after the review process. 

Application Materials: CV/Resume and a brief Statement of Interest describing relevant skills, prior experience, motivation for the project, and expected contribution.

How to Apply: Use the "Apply for this Project" button below and submit the required materials using Project Code IRP-04.

Apply for this Project

Overview: This project investigates systematic methods for evaluating the capabilities, limitations, reliability, and robustness of modern language models. The research may explore benchmark design, evaluation metrics, stress testing, adversarial and perturbation-based evaluation, robustness across domains and languages, model comparison, reproducibility, bias and failure analysis, or the development of new datasets and evaluation frameworks. Depending on the researcher's background and interests, the project may focus on methodological development, empirical studies, benchmark or dataset construction, or rigorous evaluation of existing and emerging language models.

Project Code: IRP-05

Research Areas: Benchmarking  .  Robustness  .  Reliability  .  Evaluation Metrics  .  Language Model Evaluation  .  Natural Language Processing

Student Level: Undergraduate or Graduate

Typical Duration: 4-6 months

Minimum Qualifications: Programming experience in Python; understanding of machine learning, NLP, or language models; ability to read and summarize research papers; familiarity with experimental design and data analysis; ability to work independently and meet project milestones.

Preferred Qualifications: Prior experience with LLMs, NLP libraries, evaluation frameworks, statistical analysis, dataset development, or machine learning experiments; familiarity with PyTorch, Hugging Face, or related tools; prior research or project experience in AI or NLP.

Expected Outcome: Research report and, when results warrant, preparation of a conference or journal manuscript.

Status: Open to interested students

Application Process: Interested students should submit an application using Project Code IRP-05. Applications are reviewed on a rolling basis based on relevant background, technical preparation, research interests, availability, and overall fit with the project. Selected applicants will be contacted after the review process. 

Application Materials: CV/Resume and a brief Statement of Interest describing relevant skills, prior experience, motivation for the project, and expected contribution.

How to Apply: Use the "Apply for this Project" button below and submit the required materials using Project Code IRP-05.

Apply for this Project

Overview: This project investigates systematic methods for designing language technologies inwhich humans ans AI systems work collaboratively to solve language intensive tasks. The research may explore interactive language models, human-in-the-loop learning and evaluation, adaptive AI assistance, collaborative writing and translation, decision support, explainability and user feedback, or methods for improving the effectiveness, reliability, and usability of human-AI interaction. Depending on the researcher's background and interests, the project may focus on methodological development, empirical evaluation, user-centered experimentation, benchmark or dataset construction, or domain-specific applications of collaborative language systems.

Project Code: IRP-06

Research Areas: Human-AI Collaboration  .  Interactive AI  .  Human-in-the-Loop Systems  .  Evaluation Metrics  .  Language Models  .  Natural Language Processing  .  Human-Centered AI

Student Level: Undergraduate or Graduate

Typical Duration: 4-6 months

Minimum Qualifications: Programming experience in Python; understanding of machine learning, NLP, or language models; ability to read and summarize research papers; ability to work independently and meet project milestones.

Preferred Qualifications: Prior experience with LLMs, NLP libraries, human-computer interaction, or experimental evaluation; familiarity with prompt-based systems, user studies, data analysis, or human-in-the-loop methods; experience with PyTorch, Hugging Face, or related tools; prior research or project experience in AI, NLP, HCI, or language technologies.

Expected Outcome: Research report and, when results warrant, preparation of a conference or journal manuscript.

Status: Open to interested students

Application Process: Interested students should submit an application using Project Code IRP-06. Applications are reviewed on a rolling basis based on relevant background, technical preparation, research interests, availability, and overall fit with the project. Selected applicants will be contacted after the review process. 

Application Materials: CV/Resume and a brief Statement of Interest describing relevant skills, prior experience, motivation for the project, and expected contribution.

How to Apply: Use the "Apply for this Project" button below and submit the required materials using Project Code IRP-06.

Apply for this Project