Search Solutions and Tutorials
Search Solutions is the BCS Information Retrieval Specialist Group’s annual event focused on practitioner issues in the arena of search and information retrieval. It is a unique opportunity to bring together academic research and practitioner experience.
Location: BCS London Headquarters, Ground Floor, 25 Copthall Avenue
Date: 26-27 November 2024
// our talk
What I don’t like about RAG: can we do better?
26th NOVEMBER | 11:30AM – 11:50AM (LOCAL TIME)
Retrieval Augmented Generation has been the top buzzword for the last year or so.
It consists of a first stage Retrieval (possibly followed by reranking) and a second stage Generation to Augment the retrieval with a natural language response that builds on top of the context to satisfy the information need.
It’s fascinating and promising, but all that glitters is not gold.
We are very aware that Information Retrieval is not a solved problem and building on top of it, RAG adds its own limitations: hallucinations, incorrect citations, little context influence, inaccurate results even if the context is right, wrong response format…
The scope of the talk is to explore the current techniques generally proposed to overcome such limitations and detail what is not to be liked about most of them: many presentations and implementations show sort of patched and hacky approaches rather than focusing on a reproducible and replicable solution.
The talk concludes with ideas on what we should do better and what to focus on for a more reliable future for RAG!
// our training
Retrieval Augmented Generation: Where Are We Now?
26th NOVEMBER | 1:30PM – 5:15PM (LOCAL TIME)
Retrieval Augmented Generation (RAG) is a groundbreaking approach that revolutionises how we interact with information. Imagine a powerful tool that seamlessly blends the ability to understand your questions with the capacity to retrieve and generate precise answers.
RAG leverages Information Retrieval systems (Retrieval) to build the context for Large Language Models (Augmented) and generate a natural language answer to the information need (Generation).
This tutorial explores the internals of RAG pipelines, the current limitations and promising aspects including an overview of the most interesting open-source libraries out there to build and evaluate such applications.
- 1:30-1:50 (1) Introduction to Retrieval Augmented Generation (20 mins)
- 1:50-2:30 (2) RAG Shortcomings (40 mins)
- 2:30-3:15 (3-4) RAG Solutions and Optimisation (45 mins)
- 3:15-3:30 Coffee Break (15 mins)
- 3:30-4:15 (5) Prompt Engineering Techniques and Libraries (45 mins)
- 4:15-4:45 (6) Code Example Hands-on (30 mins)
- 4:45-5:15 (7) Evaluation and wrap-up (30 mins)
// your speaker and trainer
Alessandro Benedetti
DIRECTOR @ SEASE
APACHE LUCENE/SOLR COMMITTER
APACHE SOLR PMC MEMBER
