FIRST: Faster Improved Listwise Reranking with Single Token Decoding
- Revanth Reddy
- Jaehyeok Doo
- et al.
- 2024
- EMNLP 2024
Bio
Dr. Avi Sil is a Principal Research Scientist and a people manager at IBM Research AI. He has more than 10 years of experience in the field of AI and considered a global leader in the field of Question Answering and Generative AI. He leads the Multilingual Question Answering team (comprising of research scientists and engineers) and is the creator of the PrimeQA toolkit. His team's research has not only become products but also featured at the top of multiple leaderboards (Natural Questions, TyDI, XOR-TyDI).
He is a senior program committee member for major Computational Linguistics conferences including being an Area Chair several times. He has more than 20 US patents filed all in the area of Artificial Intelligence and it's applications in various spheres.
Research Interests:
- Agentic Workflows with LLMs
- Retrieval Augmented Generation
- Large language models and representation learning
Chair:
- Demo Chair:
- NAACL 2021
- Diversity and Inclusion Chair:
- Chair of the Academic Inclusion team at ACL 2021
- Organizer:
- Question Answering and Semantic Parsing Workshop, AI Research Week, MIT 2019
- ACL Workshop on Relevance of Linguistic Structure in Neural NLP, 2018
- Shared Task Coordinator
- Area Chair (Tracks: Question Answering & Information Extraction)
- 2020: ACL, EMNLP
- 2019: CoNLL, ACL, NAACL
- 2018: NAACL, COLING
- Session Chair
- ACL 2019, Question Answering
- NAACL 2018, Question Answering
Latest News:
- (new) Invited guest lecture at GA Tech NLP course. Slides are available here.
- (new) Gave a tutorial at EMNLP 2021 Multilingual Multi-domain QA along with Sebastian Ruder. Slides are here.
- Two papers at ACL 2021 (on QA and IE) and 1 at SIGIR 2021 (Neural IR for Domain Adaptation)
- Two papers at NAACL 2021. 1. QA over tables and 2. Event time extraction
- Two papers at AAAI 2021.
- Invited talk at UCLA NLP Group (Nov 20). Host: Kai-wei Chang
Publications
Self-RAG: Learning to Retrieve, Generate, and Critique through Self-Reflection
- Akari Asai
- Zeqiu Wu
- et al.
- 2024
- ICLR 2024
Reproducible and Efficient Multi-modal Open Retrieval Question Answering
- Avi Sil
- Bhavani Iyer
- et al.
- 2023
- IJCAI 2023
Moving Beyond Downstream Task Accuracy for Information Retrieval Benchmarking
- Keshav Santhanam
- Jon Saad-Falcon
- et al.
- 2023
- ACL 2023
PrimeQA: The Prime Repository for State-of-the-Art Multilingual Question Answering Research and Development
- Avi Sil
- Jaydeep Sen
- et al.
- 2023
- ACL 2023
GAAMA 2.0: An Integrated System that Answers Boolean and Extractive Questions
- Scott McCarley
- Mihaela Bornea
- et al.
- 2023
- AAAI 2023
PrimeQA: The Prime Repository for State-of-the-Art Multilingual Question Answering Research and Development
- Avi Sil
- Jaydeep Sen
- et al.
- 2023
- arXiv
Not to Overfit or Underfit the Source Domains? An Empirical Study of Domain Generalization in Question Answering
- Arafat Sultan
- Avi Sil
- et al.
- 2022
- EMNLP 2022
Zero-Shot Dynamic Quantization for Transformer Inference
- Yousef El-Kurdi
- Jerry Quinn
- et al.
- 2022
- EMNLP 2022
Towards Robust Neural Retrieval Models with Synthetic Pre-Training
- Revanth Reddy
- Arafat Sultan
- et al.
- 2022
- SIGIR 2022
Patents
- US
- 11755657
- US
- 11651270
- US
- 11574130
- US
- 11520762
- US
- 11520829
- US
- 11010722
- US
- 10916129
- US
- 10901696
- US
- 10831800
- US
- 10691736
Blog posts
Large language models revolutionized AI. LLM agents are what’s next
ExplainerKim MartineauIBM’s AI goes multilingual — with single language training
ResearchMihaela Bornea, Lin Pan, Sara Rosenthal, Radu Florian, and Avirup Sil