Muhammed Yusuf Kocyigit

Muhammed Yusuf Kocyigit

PhD Student at Boston University

Boston University

Biography

I am currently in my fourth year of pursuing a PhD in Computer Science at Boston University, where I am being mentored by Prof. Derry Wijaya. Previouly I have worked at Amazon with Dr. Emilio Monti and at Fair(Meta) where I worked with Dr. Dieuwke Hupkes.

My research focused on developing strong cross-lingual representation learning techniques and evaluating machine translation systems. Currently, my interest lies in deepening my understanding of evaluations for Large Language Models (LLMs). This involves looking at a lot of data, examining the benchmarks used to measure LLMs generalization capabilities, and investigating potential unnoticed factors, such as contamination, that might affect our evaluation methods. Additionally, I have a keen interest in working on computational social science projects, particularly those that I believe have a significant impact.

Interests
  • Better Evaluating LLMs
  • Robust Representation Learning
  • Understanding what is in our pre-training data
  • Evaluating Machine Translation
  • Computational Social Science
Education
  • PhD in Computer Science, 2020 -

    Boston University

  • MSc in Electrical and Electronics Engineering, 2017 - 2020

    Bogazici University

  • BSc in Electrical and Electronics Engineering, 2012 -2017

    Bogazici University

Publications

(2023). A Novel Method for Analysing Racial Bias: Collection of Person Level References. Arxiv Preprint.

Cite

(2023). Western, Religious or Spiritual: An Evaluation of Moral Justification in Large Language Models. NeurIPS'23 MP2 Workshop.

Cite

(2022). AugCSE: Contrastive Sentence Embedding with Diverse Augmentations. AACL'22.

Cite

(2022). Challenges in Measuring Bias via Open-Ended Language Generation. NAACL'22 4th Workshop on Gender Bias in Natural Language Processing..

Cite

(2022). On Measuring Social Biases in Prompt-Based Multi-Task Learning. Findings of NAACL'22.

Cite

(2021). Better Quality Estimation for Low Resource Corpus Mining. Findings of ACL'22.

Cite

(2020). NUBIA: NeUral Based Interchangeability Assessor for Text Generation. INLG'20 EvalNLGEval Workshop.

Cite

(2019). Towards Neural Similarity Evaluators. NeurIPS'19 DI Workshop.

Cite

Experience

 
 
 
 
 
PhD, Research Fellow
Sep 2020 – Present Boston, MA
  • My current focus in research is learning better cross lingual representations through multitask and contrastive learning that will allow solving multiple cross lingual tasks more robustly.
  • Working on bias detection in natural language data(books and news) through time within 2021 Google Research Scholar Program.
  • Leading research project for Ottoman handwritten text recognition which received $50K from the Scientific Research Council of Turkey(TUBITAK).

Supervisor: Derry Wijaya

 
 
 
 
 
Applied Scientist Intern
Jun 2022 – Sep 2022 Cambridge, UK
  • Worked in the Semantic Understanding team. My project was focused on better evaluating and improving compositional generalization abilities of Semantic Parsing models
  • Created an extension to TMCD splits that has more robust convergence to compositionally challenging splits.
  • Worked on improving compositional generalization through meta learning and chain‑of‑thought prompting.
  • Worked quantifying characteristics of train‑test split that make a split compositionally challenging.

Supervisor: Emilio Monti

 
 
 
 
 
CEO & CoFounder
Jun 2018 – Jun 2020 Istanbul, Turkey
  • Founded company, led business development and managed client relations. Worked with the C‐Suite of LC Waikiki, multi‐billion retailer present in 40 countries, to guide their AI strategy to increase online sales.
  • Deployed an ML model for LCWaikiki for behavioral customer segmentation, customer life time value prediction and churn analysis that achieved +100% increase in conversion rates for targeted ads.
  • Worked with DEDAS, one of Turkey’s biggest electricity distributors,to detect illegal uses of electricity and increased the detection rate by +150%.
 
 
 
 
 
Technical Project Lead, ML Engineer
Turkcell Global Information
Jun 2017 – Jun 2018 Istanbul, Turkey
  • Joined AI team as a founding member. Engineered a text classification pipeline for routing customer issues.
  • Led conversational modeling project which secured a grant from Scientific Research Council of Turkey(TUBITAK).

Contact