Machine Learning Lab (MLL) | Sharif University of Technology

About the lab

Building machine learning systems that understand, reason, and generalize.

The Machine Learning Lab (MLL), directed by Dr. Mahdieh Soleymani, is a research group at Sharif University of Technology. We investigate fundamental and applied questions across modern machine learning, from robust generalization to compositional intelligence.

Generalization

Learning systems that remain reliable beyond their training distributions.

Compositional Learning

Understanding and combining concepts, attributes, and relations.

Multimodal Intelligence

Studying vision-language models, LVLMs, and reinforcement learning agents.

Generative Models

Improving control, faithfulness, and reasoning in generative systems.

Research

Questions that drive our work

We study how learning systems represent knowledge, reason over complex inputs, and remain dependable when the world differs from their training data.

Out-of-Distribution GeneralizationBuilding models that generalize beyond training distributions.

Robustness to Spurious Correlation

We develop frameworks that mitigate shortcut learning and improve robustness under subpopulation shifts, including realistic settings where spurious correlations are unknown or unlabeled.

Language-Informed Reinforcement Learning

We study how language abstractions and compositional representations can guide agents toward robust performance in unseen environments.

Compositional and Complex OoD

We create benchmarks and learning methods for shifts formed by novel compositions of previously seen concepts.

Large Vision-Language ModelsCompositional understanding and visual reasoning.

Visual Reasoning

We develop benchmarks and methods that probe multi-step reasoning and relational understanding in large vision-language models.

Representational Analysis

We examine how models such as CLIP represent objects, attributes, and their interactions to identify where compositionality emerges—and where it breaks.

InterpretabilityTracing and explaining computations inside modern models.

Mechanistic Interpretability

We develop intervention-based tools to trace how semantic functions form and transfer across layers in language and vision-language models.

Vision Transformer Attribution

We introduce more stable and faithful attribution techniques that address known failure modes in transformer explanations.

Large Language ModelsReasoning, agency, safety, and scientific discovery.

Persian Benchmarks

High-quality evaluation resources for reasoning and generation in a low-resource language.

Reasoning

Methods for more systematic, compositional, and logically consistent multi-step reasoning.

LLM Agents

Planning, memory, tool use, and coordination in long-horizon environments.

Scientific Discovery

Language models as tools for idea generation, pattern discovery, and scientific reasoning.

Chain-of-Thought Faithfulness

Evaluation of deceptive or misleading reasoning behavior in language models.

Compositional Text-to-Image GenerationFaithfully composing objects, attributes, and relations.

We design test-time strategies that improve compositional control in text-to-image generation. Our work addresses missing objects, attribute binding, and relational failures by enforcing stronger alignment between textual descriptions and visual elements.

See how these research directions translate into recent papers.

Browse publications by topic

Our community

People

Researchers and students working across the lab's core areas.

Director

Dr. Mahdieh Soleymani

Hosein Hasani

Bayesian Reinforcement Learning in Non-Stationary Environments

Mohammad Mahdi Samiei

Compositional Inductive Biases for Out of Distribution Generalization

Negin Hashemi Dijujin

Out-of-Distribution (OoD) Generalization in RL

Mohammad Hossein Narimani

Soroush Vafaie Tabar

Multiple Trigger Attacks in Backdoor Learning

Ali Bababeig

Detecting Biases in Multimodal Text–Image Foundation Models

Ali Rahimiakbar

Improving the faithfulness of reasoning in multimodal large language models.

Mohammad Mahdi Vahedi

Analysis and Enhancement of Compositional Generalization in Transformers

Fatemeh Askari

Improving Reasoning in Large Vision-Language Models

Mohammadreza Mohammadzadeh Asl

Nima Niroumand

Fatemeh Hadizadeh

Adeleh Bitarafan

Weakly Supervised Learning in 3D Medical Image Segmentation

Mahsa Ghorbani

Learning Node Embedding in Complex Networks

Fatemeh Seyedsalehi

Deep Learning in a Structured Output Space

Faezeh Faez

Deep Conditional Graph Generation for Network Completion

Seyed Mohammad Hadi Hosseini

Uncertainty-based Reasoning in Large Language Models

Arash Marioriyad

Compositional Generation in Text-to-image Models

Mohammad Mozafari

Meta-learning in 3D medical image segmentation

Ali Abdollahi

Compositional Generalization in Vision-Language Models

Alireza Roshanzamir

Alzheimer’s Disease Diagnosis using Description Test

Alireza Sahaf Naeini

Deep Networks in Reinforcement Learning

AmirHossein Ameli Kalkhoran

Continual learning using unsupervised data

AmirShayan Haghipour

Using deep generative models for event sequence generation in recommender systems

Amir Ali Moinfar

Single-Cell RNA-seq Dropout Imputation and Noise Reduction by Machine Learning

Amir Akbarnejad

Probabilistic Approach for Multi-label Classification

Danial Alihosseini

Conditional Text Generation with Deep Generative Models

Ehsan Montahaei

Adversarial Networks for Sequence Generation

Fahimeh Hosseini Noohdani

Out-of-Distribution Generalization of Image Classifiers

Faridoun Mehri

Balancing Gradient Flow for Universally Better Transformer Attributions

Fatemeh Farahnak-Ghazani

Multi-label Classification by Considering Label Dependencies

Hossein Khalili

3D Medical Images Segmentation by Effective Utilization of Unlabeled Data

Mahdi Ghaznavi

Robust Learning to Spurious Correlation without Access to Group Annotation

Marzieh Gheisari

Unsupervised Domain Adaptation

Melika Behjati

Adversarial Robustness of Deep Neural Networks in Text Domain

Amin Banayeeanzade

Meta-Continual Learning

Mohamadreza Fereydooni

Many-Class Few-Shot Classification

Omid Abbasi

Deep Learning in Recommender Systems

Parishad BehnamGhader

Graph-based Word Embedding using Deep Neural Networks

Rasool Mirzaiezadeh

Few-Shot Semantic Segmentation using Meta-Learning

Sarah Rastegar

Deep Multi-Modal Learning

Seyed Alireza Mirmohammad Sadeghi

Deep Video Captioning using Recurrent Neural Networks

Roostaiyan Seyed Mahdi

Multi-modal Distance Metric Learning

Seyed Mohammad Chavoshian

Deep Visual Question Answering

Seyed Mohsen Shojaee

Deep Zero-Shot Learning

Seyed Roozbeh Razavi Rohani

Meta Reinforcement Learning using Brain-Inspired Networks

Sina Hajimiri

Representation Learning by Deep Networks and Information Theory

Zeinab Golgooni

Deep Learning Based Proarrhythmia Analysis

Teaching

Courses

Selected courses taught by members of the lab at Sharif University of Technology.

Fall 2025

Artificial Intelligence

Visit course website

Spring 2025

System II

Visit course website

Spring 2024

Modern Information Retrieval

Website unavailable

Spring 2024

Deep Learning

Website unavailable

Fall 2023

Large Language Models

Visit course website

2022