BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Memento EPFL//
BEGIN:VEVENT
SUMMARY:NLP seminar: Retrieving Texts based on Abstract Descriptions
DTSTART:20231121T110000
DTEND:20231121T120000
DTSTAMP:20260510T013133Z
UID:2d92d04e9ecf3a318227326d20bc6414f65ba92757dd57574d67fb5b
CATEGORIES:Conferences - Seminars
DESCRIPTION:Shauli Ravfogel (Bar-Ilan University)  \nShauli Ravfogel\, f
 rom Bar-Ilan University\, is presenting his most recent work on Retrievin
 g Texts based on Abstract Descriptions. \nYou can join in BC 04 or onlin
 e.\n\nThis talk aims to connect two research areas: instruction models and
  retrieval-based models.\n\nAbstract:\nWhile instruction-tuned Large Langu
 age Models (LLMs) excel at extracting information from text\, they are not
  suitable for semantic retrieval. Similarity search over embedding vectors
  allows to index and query vectors\, but the similarity reflected in the e
 mbedding is sub-optimal for many use cases. We identify the task of retrie
 ving sentences based on abstract descriptions of their content.\nWe demons
 trate the inadequacy of current text embeddings and propose an alternative
  model that significantly improves when used in standard nearest neighbor 
 search. The model is trained using positive and negative pairs sourced thr
 ough prompting a large language model (LLM). While it is easy to source th
 e training material from an LLM\, the retrieval task cannot be performed b
 y the LLM directly. This demonstrates that data from LLMs can be used not 
 only for distilling more efficient specialized models than the original LL
 M\, but also for creating new capabilities not immediately possible using 
 the original model.
LOCATION:BC 04 https://plan.epfl.ch/?room==BC%2004 https://epfl.zoom.us/j/
 69499602273?pwd=WTBWK1o0L1Z0b085anBiM094STFjQT09
STATUS:CONFIRMED
END:VEVENT
END:VCALENDAR