Nguyen Hoang Thao
0 followers · 231 views
on the atlas — 30
- State as a Snapshot – React2 savers
- ĐĂNG KÝ DỰ TUYỂN ĐÀO TẠO THẠC SĨ1 savers
- Optimizing RAG with Hybrid Search & Reranking | VectorHub by Superlinked1 savers
- Introducing Hybrid Search and Rerank to Improve the Retrieval Accuracy of the RAG System - Dify Blog1 savers
- Rerankers and Two-Stage Retrieval | Pinecone2 savers
- Tree of Thoughts (ToT) | Prompt Engineering Guide1 savers
- Prompt Chaining | Prompt Engineering Guide1 savers
- Generated Knowledge Prompting | Prompt Engineering Guide1 savers
- Self-Consistency | Prompt Engineering Guide2 savers
- Chain-of-Thought Prompting | Prompt Engineering Guide1 savers
- Few-Shot Prompting | Prompt Engineering Guide2 savers
- Zero-Shot Prompting | Prompt Engineering Guide1 savers
- LLM Settings | Prompt Engineering Guide2 savers
- Prompt Engineering Guide | Prompt Engineering Guide7 savers
- From zero to hero with LLMs1 savers
- What is a SEMANTIC Knowledge Graph?1 savers
- Revolutionizing Semantic Search with RAG and Knowledge Graphs1 savers
- Introduction to Large Language Models | Machine Learning | Google for Developers2 savers
- LCh-015B.pdf1 savers
- RAG Architecture Deep Dive | LinkedIn1 savers
- RAG and generative AI - Azure AI Search | Microsoft Learn1 savers
- What is Retrieval Augmented Generation (RAG)? | Databricks1 savers
- arXiv:1409.3215v3 [cs.CL] 14 Dec 20141 savers
- C:\Documents and Settings\nigel\デスクトップ\Cover pages.prn.pdf1 savers
- smbm16Hahn.pdf1 savers
- Structured Streaming Programming Guide - Spark 3.5.0 Documentation1 savers
- 1612.03284.pdf1 savers
- Updating Arrays in State – React1 savers
- Updating Objects in State – React1 savers
- Queueing a Series of State Updates – React1 savers
highlights — 320
But these top-k matches may not always include the relevant sections
Optimizing RAG with Hybrid Search & Reranking | VectorHub by SuperlinkedReciprocal Rank Fusion (RRF) is one of several available methods for combining dense and sparse search scores
Optimizing RAG with Hybrid Search & Reranking | VectorHub by SuperlinkedThe hybrid score is a pure vector score when α is 1, and a pure keyword score when α is 0.
Optimizing RAG with Hybrid Search & Reranking | VectorHub by SuperlinkedA common approach to vector search is cosine similarity search
Optimizing RAG with Hybrid Search & Reranking | VectorHub by Superlinkedis a refined version of the TF-IDF
Optimizing RAG with Hybrid Search & Reranking | VectorHub by Superlinkedsparse vector search
Optimizing RAG with Hybrid Search & Reranking | VectorHub by SuperlinkedBM25 (Best Match 25)
Optimizing RAG with Hybrid Search & Reranking | VectorHub by Superlinkednot as a replacement for search technology, but as a supplementary tool
Introducing Hybrid Search and Rerank to Improve the Retrieval Accuracy of the RAG System - Dify BlogIt's also beneficial in a single search mode
Introducing Hybrid Search and Rerank to Improve the Retrieval Accuracy of the RAG System - Dify Blognecessary to select a suitable segmentation strategy and TopK value that align with the model's input window size constraints
Introducing Hybrid Search and Rerank to Improve the Retrieval Accuracy of the RAG System - Dify Bloga scoring system, specifically the Rerank Model, becomes essential
Introducing Hybrid Search and Rerank to Improve the Retrieval Accuracy of the RAG System - Dify Blogit's necessary to merge and normalize query results from different search modes
Introducing Hybrid Search and Rerank to Improve the Retrieval Accuracy of the RAG System - Dify BlogPopular rerank models include Cohere rerank, bge-reranker,
Introducing Hybrid Search and Rerank to Improve the Retrieval Accuracy of the RAG System - Dify Blogknowledge graph technology
Introducing Hybrid Search and Rerank to Improve the Retrieval Accuracy of the RAG System - Dify BlogThis article uses the integration of vector and keyword retrieval
Introducing Hybrid Search and Rerank to Improve the Retrieval Accuracy of the RAG System - Dify Blogunderstanding your users and the types of questions they are most likely to ask is essential.
Introducing Hybrid Search and Rerank to Improve the Retrieval Accuracy of the RAG System - Dify Blogrerankers are slow, and retrievers are fast
Rerankers and Two-Stage Retrieval | Pineconea type of model that, given a query and document pair, will output a similarity score
Rerankers and Two-Stage Retrieval | PineconeThe solution to this issue is to maximize retrieval recall by retrieving plenty of documents and then maximize LLM recall by minimizing the number of documents that make it to the LLM
Rerankers and Two-Stage Retrieval | PineconeResearch shows that LLM recall degrades as we put more tokens in the context window
Rerankers and Two-Stage Retrieval | PineconeWhen storing information in the middle of a context window, an LLM's ability to recall that information becomes worse than had it not been provided in the first place [2]
Rerankers and Two-Stage Retrieval | Pineconewe call this limit the context window
Rerankers and Two-Stage Retrieval | Pineconelimits on how much text we can pass to them
Rerankers and Two-Stage Retrieval | Pineconerecall — meaning "how many of the relevant documents are we retrieving"
Rerankers and Two-Stage Retrieval | Pineconewe cannot return everything
Rerankers and Two-Stage Retrieval | Pineconethe retrieval may return relevant information below our top_k cutoff
Rerankers and Two-Stage Retrieval | PineconeThere is some information loss because we're compressing this information into a single vector.
Rerankers and Two-Stage Retrieval | Pineconeomplex problem solving through tree search via a multi-round conversation
Tree of Thoughts (ToT) | Prompt Engineering Guidedifferent tasks requires defining the number of candidates and the number of thoughts/steps.
Tree of Thoughts (ToT) | Prompt Engineering GuideTree of Thoughts (ToT), a framework that generalizes over chain-of-thought prompting and encourages exploration over thoughts that serve as intermediate steps
Tree of Thoughts (ToT) | Prompt Engineering GuideThis is what's referred to as prompt chaining, where a task is split into subtasks with the idea to create a chain of prompt operations.
Prompt Chaining | Prompt Engineering Guideincorporate knowledge or information to help the model make more accurate predictions.
Generated Knowledge Prompting | Prompt Engineering Guideself-consistency aims "to replace the naive greedy decoding used in chain-of-thought prompting"
Self-Consistency | Prompt Engineering Guideeliminate manual efforts by leveraging LLMs with "Let's think step by step" prompt to generate reasoning chains for demonstrations one by one
Chain-of-Thought Prompting | Prompt Engineering Guidechain-of-thought (CoT) prompting enables complex reasoning capabilities through intermediate reasoning steps
Chain-of-Thought Prompting | Prompt Engineering Guideis still not a perfect technique, especially when dealing with more complex reasoning tasks
Few-Shot Prompting | Prompt Engineering Guidethe format you use also plays a key role
Few-Shot Prompting | Prompt Engineering GuideInstruction tuning is essentially the concept of finetuning models on datasets described via instructions
Zero-Shot Prompting | Prompt Engineering GuideZero-shot prompting means that the prompt used to interact with the model won't contain examples or demonstrations
Zero-Shot Prompting | Prompt Engineering Guidealter the frequency or presence penalty but not both.
LLM Settings | Prompt Engineering GuideA token that appears twice and a token that appears 10 times are penalized the same
LLM Settings | Prompt Engineering GuideThis setting reduces the repetition of words in the model's response by giving tokens that appear more a higher penalty
LLM Settings | Prompt Engineering Guidea string that stops the model from generating tokens
LLM Settings | Prompt Engineering Guidealter temperature or Top P but not both
LLM Settings | Prompt Engineering GuideIf you use Top P it means that only the tokens comprising the top_p probability mass are considered for responses
LLM Settings | Prompt Engineering GuideA sampling technique with temperature, called nucleus sampling
LLM Settings | Prompt Engineering Guidepoem generation or other creative tasks, it might be beneficial to increase the temperature value
LLM Settings | Prompt Engineering Guidea lower temperature value for tasks like fact-based QA
LLM Settings | Prompt Engineering GuideIncreasing temperature could lead to more randomness, which encourages more diverse or creative outputs
LLM Settings | Prompt Engineering GuideWhen designing and testing prompts, you typically interact with the LLM via an API
LLM Settings | Prompt Engineering Guide