Suggestions
Paolo Papotti
Associate Professor at EURECOM
Paolo Papotti is an Associate Professor in the Data Science department at EURECOM, a research and higher education institution located in France. He has held this position since 2017 after previously working as a scientist in the data analytics group at the Qatar Computing Research Institute (QCRI) and as an assistant professor at Arizona State University in the USA. Papotti earned his Ph.D. from the University of Roma Tre in Italy in 2007.132
Research Focus
Papotti's research interests lie primarily in data management, particularly in areas such as:
- Data integration
- Information quality
- User-centric data cleaning
- Automatic fact-checking
He has made significant contributions to the field, with numerous publications addressing topics like schema mapping, data cleaning systems, and scalable data management techniques.34
Professional Contributions
In addition to his teaching responsibilities, he serves on several editorial boards and program committees for major conferences in his field, such as SIGMOD and VLDB. His work has garnered considerable recognition, as indicated by his citation metrics, marking him as an influential figure in data science.34
Contact Information
For academic inquiries or collaboration opportunities, Paolo Papotti can be contacted via email at [email protected].1
Highlights
Forget monster GPUs requirements on long context scenarios! FINCH compresses LLMs' context by up to 93x, allowing users to fit a book into a LLM with modest hardware requirements! 🚀✨
FINCH is a novel method designed to compress input context by selectively storing only the most pertinent Key-Value pairs in the cache. This innovation allows decoder-based LLMs to handle longer texts more efficiently without the need for fine-tuning.
How does it work? 🛠️ By leveraging the pre-trained weights of the self-attention mechanism, FINCH identifies the most relevant Key (K) and Value (V) pairs, significantly reducing the memory footprint while maintaining output quality.
Performance Highlights 📈 FINCH delivers compression ratios of up to 93x, ensuring comparable generation quality while being faster and more memory-efficient than traditional methods.
✨ Key Results: On benchmarks like SQuAD v2 and LongBench, FINCH consistently excels in tasks such as question answering, summarization, and code completion, demonstrating high accuracy and faster execution times.
Plug-and-Play Solution 🔌 One of FINCH's features is its seamless integration with any decoder-based model. There's no need for retraining or fine-tuning—just specify your target compression size, and FINCH dynamically optimizes for you. This makes it easy to tap into more efficient LLMs without the usual hassles.
Why It Matters 📊 Large Language Models, like ChatGPT, often face GPU memory constraints and they are limited by the context window defined during their pretraining. FINCH addresses this challenges by optimizing resource usage and let the long context fit within the model context’s window.
Paper: https://lnkd.in/d4XKtTQ7
Code repository on GitHub: https://lnkd.in/d5eM4r7g
Our compression method will be featured in Transactions on ACL (#TACL) and showcased at #EMNLP2024. This work marks the first milestone in Giulio Corallo's Ph.D. journey at SAP under the CIFRE thesis program. 📚🔬
#AI #NLP #Transformers
