I am a Ph.D. researcher in IEMS at the University of Central Florida's Complex Adaptive Systems Lab, advised by Dr. Niloofar Yousefi. My research is about reliable LLMs when humans can't supervise at scale. Four concrete questions drive the work. (1) Can pretrained transformers report calibrated uncertainty in their own attention? (2) Can we verify AI-generated supervision when the judges themselves share blind spots? (3) Can we monitor tool-using agents and intervene on unsafe trajectories, not just outputs? (4) Can we steer a model away from outdated parametric memory at test time, without retraining or retrieval? My applied vertical is clinical and biomedical AI, where the cost of confident errors is measured in patient harm, not benchmarks.
-
Test-Time Uncertainty in Pretrained Transformers
- Inference-time uncertainty-aware attention for pretrained transformers UAT-LITE
- Bayesian uncertainty quantification for safe clinical decision support MedBayes-Lite
-
Reliable AI Supervision
- Fault-tolerant preference alignment via multi-agent verification MAV
- Reliability-aware filtering when LLM judges share blind spots CorrFilter
- Runtime Safety and Evaluation for LLM Agents
- Test-Time Steering Without Retraining
Recent updates
Featured Publications
Check out Google Scholar for a full list of my publications.
UAT-LITE: Inference-Time Uncertainty-Aware Attention for Pretrained Transformers
Inference-time method that uses Monte-Carlo dropout to make pretrained transformer attention uncertainty-aware. Improves calibration, selective prediction, and robustness without retraining.
@article{hossain2026uatlite,
title={UAT-LITE: Inference-Time Uncertainty-Aware Attention for Pretrained Transformers},
author={Hossain, Elias and Dipta, Shubhashis Roy and Neupane, Subash and Rana, Rajib and Shwartz-Ziv, Ravid and Garibay, Ivan},
journal={arXiv preprint arXiv:2602.02952},
year={2026}
}
Computational Paradigms for Antimicrobial Resistance Prediction: Integrating Multi-Omics, Structural Modeling, and Foundation AI Systems
Survey unifying multi-omics, structural modeling, and foundation models for AMR prediction; positions where each paradigm shines and where they break.
@article{hossain2026computational,
title={Computational paradigms for antimicrobial resistance prediction: integrating multi-omics, structural modeling, and foundation artificial intelligence systems},
author={Hossain, Elias and Yousefi, Niloofar},
journal={Briefings in Bioinformatics},
volume={27},
number={3},
pages={bbag219},
year={2026},
publisher={Oxford University Press}
}
BIOGEN: Evidence-Grounded Multi-Agent Reasoning Framework for Transcriptomic Interpretation in Antimicrobial Resistance
A multi-agent framework that interprets RNA-Seq gene clusters by combining biomedical retrieval with evidence verification; every cluster interpretation cites at least one PMID, DOI, or UniProt accession across five bacterial datasets.
@article{hossain2026biogen,
title={BIOGEN: Evidence-Grounded Multi-Agent Reasoning Framework for Transcriptomic Interpretation in Antimicrobial Resistance},
author={Hossain, Elias and Shoeibi, Mehrdad and Garibay, Ivan and Yousefi, Niloofar},
journal={Frontiers in Bioinformatics},
volume={6},
year={2026},
publisher={Frontiers Media}
}
When Policies Cannot Be Retrained: A Unified Closed-Form View of Post-Training Steering in Offline Reinforcement Learning
A closed-form analysis of post-training policy steering in offline RL, showing when interventions help, hurt, or leave a frozen policy unchanged.
@article{hossain2026policies,
title={When Policies Cannot Be Retrained: A Unified Closed-Form View of Post-Training Steering in Offline Reinforcement Learning},
author={Hossain, Elias and Basher, M. J. I. and Garibay, Ivan and Garibay, Ozlem and Yousefi, Niloofar},
journal={arXiv preprint arXiv:2604.22873},
year={2026}
}
R-GAT: Cancer Document Classification Leveraging Graph-Based Residual Network for Scenarios with Limited Data
Residual graph attention network for biomedical document classification under limited data; matches transformer baselines on cancer abstracts at a fraction of the cost.
@article{hossain2026rgat,
title={R-GAT: cancer document classification leveraging graph-based residual network for scenarios with limited data},
author={Hossain, Elias and Nuzhat, Tasfia and Masum, Shamsul and Rahimi, Shahram and Golilarz, Noorbakhsh Amiri},
journal={Scientific Reports},
year={2026},
publisher={Nature Publishing Group}
}
From Questions to Insightful Answers: Building an Informed Chatbot for University Resources
Retrieval-augmented, domain-informed chatbot framework for university support; demonstrates effective conversational agents on grounded institutional knowledge.
@inproceedings{neupane2024questions,
title={From Questions to Insightful Answers: Building an Informed Chatbot for University Resources},
author={Neupane, Subash and Hossain, Elias and Keith, Jason and Tripathi, Himanshu and Ghiasi, Farbod and Golilarz, Noorbakhsh Amiri and Amirlatifi, Amin and Mittal, Sudip and Rahimi, Shahram},
booktitle={2024 IEEE Frontiers in Education Conference (FIE)},
year={2024},
organization={IEEE}
}