Praveen Bushipaka

About

I am a PhD researcher in Artificial Intelligence, with the past four years of my work focused mainly on NLP and Large Language Models (LLMs). I am currently pursuing my PhD under the supervision of Prof. Tommaso Cucinotta and Dr. Lucia Passaro.

My research focuses on LLM unlearning, data selection, trustworthy AI, and retrieval. In particular, I work on gradient-guided data selection, reliable evaluation of machine unlearning, and privacy-preserving ways of adapting language models.I have also worked on multi-perspective learning in Social AI, exploring how different human perspectives and opinions can be represented using soft labels rather than reducing them to a single ground-truth label.

Beyond research, I enjoy building practical AI systems. My engineering experience includes distributed training, parameter-efficient fine-tuning (PEFT), retrieval-augmented generation (RAG), evaluation pipelines, and full-stack AI tools.

Before moving into NLP and LLM research, I worked with traditional machine learning in industry, where I developed and deployed ML models for real-world applications. This combination of research and engineering experience has shaped my interest in taking new NLP and LLM ideas beyond experiments and turning them into reliable, measurable systems.

Research Interests

  • Large Language Models
  • Machine Unlearning
  • Data Selection & Coresets
  • Trustworthy AI
  • Retrieval-Augmented Generation
  • LLM Systems & Evaluation

Selected Publications

2026

GRACE: Gradient-guided Coreset Selection for LLM Unlearning

Praveen Bushipaka, Andrea D’Angelo, Lucia Passaro, Tommaso Cucinotta

Findings of EMNLP, 2026Accepted

A gradient-guided coreset selection method for constructing compact forget and retain sets for LLM unlearning.

arXiv
Cite 
@misc{bushipaka2026gracegradientguidedcoresetselectionllm,
  title={GRACE: Gradient-guided Coreset Selection for LLM Unlearning},
  author={Praveen Bushipaka and Andrea D'Angelo and Lucia Passaro and Tommaso Cucinotta},
  year={2026},
  eprint={2608.28361},
  archivePrefix={arXiv},
  primaryClass={cs.AI},
  url={https://arxiv.org/abs/2608.28361}
}

OARS: Objective-Aware Retain-Set Selection for LLM Unlearning

Praveen Bushipaka, et al.

NAACL, 2026Under Review

Data-centric research on reliable LLM unlearning.

ForgetLLM: A Practical Implementation of LLM Unlearning

Praveen Bushipaka, et al.

AAAI Demo, 2026Under Review

A full-stack platform for modular LLM unlearning, data selection, distributed training, and experiment evaluation.

2025

Standard vs. Modular Sampling: Best Practices for Reliable LLM Unlearning

Praveen Bushipaka, Lucia Passaro, Tommaso Cucinotta

WIPE-OUT Workshop at ECML PKDD 2025 · Springer CCIS, vol. 2840 (2026), 2025

Paper
Cite 
@incollection{bushipaka2026standard,
  title={Standard vs. Modular Sampling: Best Practices for Reliable LLM Unlearning},
  author={Bushipaka, Praveen and Passaro, Lucia and Cucinotta, Tommaso},
  booktitle={Machine Learning and Principles and Practice of Knowledge Discovery in Databases},
  editor={Koprinska, Irena and Mendes-Moreira, João and Branco, Paula},
  series={Communications in Computer and Information Science},
  volume={2840},
  publisher={Springer},
  address={Cham},
  year={2026},
  doi={10.1007/978-3-032-19099-4_43},
  url={https://doi.org/10.1007/978-3-032-19099-4_43}
}

Embracing Diversity: A Multi-Perspective Approach with Soft Labels

Benedetto Muscato, Praveen Bushipaka, Gizem Gezici, Lucia PASS Aro, Fosca Giannotti, Tommaso Cucinotta

HHAI 2025: Proceedings of the 4th International Conference on Hybrid Human-Artificial Intelligence, 2025, pp. 370–383

Paper
Cite 
@inproceedings{muscato2025embracing,
  title={Embracing Diversity: A Multi-Perspective Approach with Soft Labels},
  author={Muscato, Benedetto and Bushipaka, Praveen and Gezici, Gizem and Aro, Lucia PASS and Giannotti, Fosca and Cucinotta, Tommaso},
  booktitle={HHAI 2025: Proceedings of the 4th International Conference on Hybrid Human-Artificial Intelligence},
  pages={370--383},
  year={2025},
  doi={10.3233/FAIA250654},
  url={https://journals.sagepub.com/doi/10.3233/FAIA250654},
  organization={SAGE Publications 1 Oliver's Yard, 55 City Road, London, EC1Y 1SP}
}

2024

A Novel Multi-Step Prompt Approach for LLM-based Q&As on Banking Supervisory Regulation

Daniele Licari, Canio Benedetto, Praveen Bushipaka, Alessandro De Gregorio, Marco De Leonardis, Tommaso Cucinotta

Proceedings of the Tenth Italian Conference on Computational Linguistics (CLiC-it 2024), 2024, pp. 496–509

A multi-step prompt construction method for more precise and informative answers about banking supervisory reporting obligations.

Paper
Cite 
@inproceedings{licari-etal-2024-novel,
  title={A Novel Multi-Step Prompt Approach for {LLM}-based {Q}{&}As on Banking Supervisory Regulation},
  author={Licari, Daniele and Benedetto, Canio and Bushipaka, Praveen and De Gregorio, Alessandro and De Leonardis, Marco and Cucinotta, Tommaso},
  editor={Dell'Orletta, Felice and Lenci, Alessandro and Montemagni, Simonetta and Sprugnoli, Rachele},
  booktitle={Proceedings of the Tenth Italian Conference on Computational Linguistics (CLiC-it 2024)},
  month={December},
  year={2024},
  address={Pisa, Italy},
  publisher={CEUR Workshop Proceedings},
  url={https://aclanthology.org/2024.clicit-1.59/},
  pages={496--509},
  isbn={979-12-210-7060-6}
}

2023

Legal Holding Extraction from Italian Case Documents using Italian-LEGAL-BERT Text Summarization

Daniele Licari, Praveen Bushipaka, Gabriele Marino, Giovanni Comandé, Tommaso Cucinotta

Proceedings of the Nineteenth International Conference on Artificial Intelligence and Law (ICAIL ’23), 2023, pp. 148–156

Paper
Cite 
@inproceedings{10.1145/3594536.3595177,
  author={Licari, Daniele and Bushipaka, Praveen and Marino, Gabriele and Comand{\'e}, Giovanni and Cucinotta, Tommaso},
  title={Legal Holding Extraction from Italian Case Documents using Italian-LEGAL-BERT Text Summarization},
  year={2023},
  isbn={9798400701979},
  publisher={Association for Computing Machinery},
  address={New York, NY, USA},
  url={https://doi.org/10.1145/3594536.3595177},
  doi={10.1145/3594536.3595177},
  booktitle={Proceedings of the Nineteenth International Conference on Artificial Intelligence and Law},
  pages={148--156},
  numpages={9},
  keywords={Benchmark Dataset, Extractive Text Summarization, Holding Extraction, Italian-LEGAL-BERT},
  location={Braga, Portugal},
  series={ICAIL '23}
}