Ministry of peace

Leeds united

Retrieval Augmented Generation (RAG) is one of the most popular techniques to improve the accuracy and reliability of Large Language Models (LLMs). This is possible by providing additional information from external data sources. This way, the answers can be tailored to specific contexts and updated without fine-tuning and retraining.

Despite the popularity of RAG, it doesn’t work well in all contexts. In this article, we’ll explain step-by-step how to build RAG pipelines effectively, explore the limitations of RAG, and how to address these limitations.

seeing the power of math in this video, I remembered smth. when we were in high school almost we all asked "why do we learn math?" did u any of u say 'we learn it to boost our thinking skill and problem solving"? or did u say, "it is just finding the value of x and y for years that nobody has got"?😂 Yh and there’s actually a very practical answer now. A lot of us in high school asked, “Why are we learning math? Is it only solving for x and y forever?” 😂But math quietly ended up powering a lot of the technology we use tdy especially search, information retrieval, and modern AI. ena When you search for a document, the computer usually doesn’t “understand” words the way humans do. It first converts text into numbers. That process is called an embedding. An embedding turns a word, sentence, or whole document into a vector basically a list of numbers that captures meaning. lemsale, words like “university,” “student,” and “campus” often end up closer together in that mathematical space than unrelated words like “banana” or “engine.”(sill they hv similarity but it is much too low.)
Then math helps compare those vectors. One common method is cosine similarity. equatinu degmo paper lay ale
Instead of checking whether two texts use exactly the same words, cosine similarity checks whether their vectors point in similar directions. If they do, the meanings are probably related. A simplified version of the process looks like this: mejemereya You ask a question kezia The system converts your question into an embedding (numbers) after that It already has stored embeddings for documents. then It compares your question vector with document vectors. finally. The documents with the highest similarity are retrieved.
That is a big part of vector space retrieval in the field of Information Retrieval.This is also how many RAG systems work. Before the AI writes an answer, it first retrieves the most relevant chunks of information using mathematical similarity. Then the model uses those retrieved pieces as context to generate a better answer.
So the funny thing is bzu sew a lot of what looked like “just x and y” in school later becomes vectors, distances, similarity scores, optimization, and probability and that’s part of what makes search engines and modern AI work.
and honestly AAU Chat bot ena Adwa AI assistant bezi new yeseranew.

@code_it_now

▓▓▓▓▓▓▓▓▓▓▓░░░░░ ፷፭% (65%)

BoA and dashen super app has more bug than በግ ተራ

Can I Survive a Day in Addis on ETB 600?

Vibe coder era😭😭

What do u think abt this one?? i want to know ur view

I hope you’re all having a great day! I have a quick favor to ask. If you could take a moment to sign up using the link below, I would really appreciate it!

👉 AfterQuery 👈

Your support means a lot, and it will help me out tremendously. Thank you so much for your time!

Hey everyone, Abuki is here .
A full stack dev , passionate about software development, saas , blockchain, fintech and digital opportunities.

I created this channel to share my journey and move my thoughts from saved messages to here .

If you’re into tech, and you are welcome here 🫡

Welcome to my channel, everyone.🔥

@abukidev0