Samuele Poppi
Scholar

Samuele Poppi

Google Scholar ID: EBLQPgcAAAAJ
Mohamed bin Zayed University of Artificial Intelligence (MBZUAI), previously GenAI at Meta
Responsible AIMachine UnlearningLLM SecurityWatermarking
Homepage↗Google Scholar↗
Citations & Impact
All-time
Citations
187
 
H-index
6
 
i10-index
5
 
Publications
13
 
Co-authors
0
 
Contact
Emails.poppi94@gmail.comCVOpen ↗TwitterOpen ↗GitHubOpen ↗LinkedInOpen ↗
Publications
14 items
Constitution-Guided Watermarking
2026
Cited
0
Certification of Real Images through Calibrated Content Authentication
2026
Cited
0
ShieldCLIP: Selective Safety Alignment for Harmful Content Mitigation in Multimodal Foundation Models
2026
Cited
0
CopyShield: A Cross-Level Benchmark of Copyright Defenses in LLMs
2026
Cited
0
Context Inference Attacks Without Jailbreaks
2026
Cited
0
ReACT-CLIP: Response-Aware Test-Time Defense for Vision--Language Models
2026
Cited
0
A Gravitational Interpretation of Fine-Tuning Reversion
2026
Cited
0
When the Chain of Thought Knows Better: Failure Modes in Multi-Turn Reasoning Models
2026
Cited
0
Resume
Academic Achievements
  • Mitigating Watermark Forgery in Generative Models via Randomized Key Selection (2025)
  • Improving LLM First-Token Predictions in Multiple-Choice Question Answering via Output Prefilling (2025)
  • Towards Understanding the Fragility of Multilingual LLMs against Fine-Tuning Attacks (NAACL 2024)
  • Unlearning Vision Transformers without Retaining Data via Low-Rank Decompositions (ICPR 2024)
  • Safe-CLIP: Removing NSFW Concepts from Vision-and-Language Models (ECCV 2023)