Scholar
Federico Cocchi
Google Scholar ID: BRG3e1EAAAAJ
PhD student, University of Modena and Reggio Emilia
Citations & Impact
Citations
346
H-index
7
i10-index
7
Publications
13
Co-authors
10
list available
Publications
Resume (English only)
Academic Achievements
- Paper 'Augmenting Multimodal LLMs with Self-Reflective Tokens for Knowledge-based Visual Question Answering' accepted at CVPR 2025
- Paper 'The (R)Evolution of Multimodal Large Language Models: A Survey' accepted at ACL Findings 2024
- Paper 'Wiki-LLaVA: Hierarchical Retrieval-Augmented Generation for Multimodal LLMs' accepted at CVPR Workshop 2024
- Published research at top-tier conferences including CVPR, ECCV, ACL, ICCV Workshop, and ICPR
- Proposed ReflectiVA model that integrates external knowledge via reflective tokens for improved knowledge-based VQA
- Developed Wiki-LLaVA with hierarchical retrieval pipeline to augment MLLMs with external knowledge without compromising standard benchmark performance
Background
- Final-year PhD student in the Italian National PhD Program in Artificial Intelligence
- Research focuses on Multimodal LLMs, especially at the intersection of vision and language
- Aims to enhance reasoning and understanding capabilities of models
- Explores post-training techniques to enrich models with retrieval and reranking using multimodal data
- Uses HPC systems for multi-GPU/multi-node foundation model training through collaboration with CINECA
- Research interests include Generative AI, Computer Vision, and Natural Language Processing
