CV

Remote Research Collaborator

Tilburg University, Netherlands

Investigated how different language models use multiple contextual cues to disambiguate grammatical tasks.

• Language Models
• Interpretability
• Context Mixing
• Activation Patching

Remote Research Collaborator

IT University of Copenhagen, Denmark (NLPnorth lab)

Developed a more complete lang2vec system to improve multilingual models performance.

• Multilingual NLP
• Classical Machine Learning pipelines

Master Student

Sharif University of technology, Iran

Focusing on interpretability of transformer based language models

• Mechanistic Interpretability
• Transformer Models
• AI Safety

Master

Sharif University of technology

CE, Artificial Intelligence

• Mechanistic Interpretability
• Transformer Models
• AI Safety

BSc

Tehran Polytechnic (Amirkabir University of Technology)

Mechanical Engineering

• Control Engineering
• Dynamical Systems
• Signal Processing

data2lang2vec: Data Driven Typological Features Completion

arXiv preprint (COLING 2025 under review)

Developed a more complete lang2vec system to improve multilingual models performance.

How Language Models Prioritize Contextual Grammatical Cues?

BlackboxNLP @ EMNLP 2024

Investigated how different language models use multiple contextual cues to disambiguate grammatical tasks.

HalluSafe at SemEval‑2024 Task 6: An NLI‑based Approach to Make LLMs Safer by Better Detecting Hallucinations and Overgeneration Mistakes

SemEval 2024

Particpated in a shared task of SemEval 2024 titled SHROOM, a Shared‑task on Hallucinations and Related Observable Overgeneration Mistakes of LMs.

Imaginations of WALL‑E : Reconstructing Experiences with an Imagination‑Inspired Module for Advanced AI Systems

arXiv preprint

Proposed an imagination inspired module to enhance the performance of multi‑modal LLMs on text only tasks.