Erika Shorten (@eshorten300) / X

Erika Shorten

4,361 posts

Erika Shorten

@eshorten300

Partnerships @weaviate_io | Diary about agents, LLM frameworks, and vector databases 🤪

Boston

github.com/weaviate/recip…

Joined August 2019

Erika Shorten
@eshorten300
Nov 25, 2024
I'm excited to share my talk on building Agentic RAG systems at last week's event with @arizeai and @Google! 🥳 My talk covered: 1. The differences between Vanilla RAG and Agentic RAG 2. The agent ecosystem and how you can build agents today 3. How @weaviate_io is building
00:00
61K
Erika Shorten
@eshorten300
Nov 19, 2023
RAG evaluation is the next step once your pipeline is set ☝️ There are four new metrics based on LLM evaluation that are setting the standard: 1. Faithfulness, 2. Answer relevancy, 3. Context precision, and 4. Context recall (Ragas score) The visual below illustrates prompting
132K
Erika Shorten
@eshorten300
Dec 21, 2023
There is a lot of discussion around building a search engine rather than a vector database. There is more to search than retrieving relevant documents. Your RAG pipeline could benefit from hybrid search and reranking ☝️ RankZephyr is an open-source LLM for zero-shot listwise
84K
Erika Shorten
@eshorten300
Oct 15, 2023
Build a RAG Application from Scratch 🐕 This video covers the architecture behind Verba 0.3.0! The design behind Verba is to have an explicit manager for each core component of RAG pipelines (Read, Chunk, Embed, Retrieve, Generate) 1. ReaderManager: Load in your data
00:00
155K
Erika Shorten
@eshorten300
Nov 6, 2024
Swapping out RAG for Agentic RAG Systems 🤖 Agents differ from the standard retrieve and generate because of their access to tools, memory, and planning. Why is this game-changing? It allows us to build systems with complete autonomy to reason and execute specific tools when
Leonie
@helloiamleonie
Nov 5, 2024
Goodbye, vanilla RAG. Hello, Agentic RAG! 𝗩𝗮𝗻𝗶𝗹𝗹𝗮 𝗥𝗔𝗚 The common vanilla RAG implementation processed the user query through a retrieval and generation pipeline to generate a response grounded in external knowledge. Advanced vanilla RAG techniques include e.g.,
56K
Erika Shorten
@eshorten300
Oct 22, 2023
I'm in the process of learning about Agents 🤖 @lilianweng wrote a great post covering the key components of an LLM-powered autonomous agent system. Here is a short overview of how each piece builds on an LLM: • Planning: The agent can break down complex tasks into subtasks.
52K
Erika Shorten
@eshorten300
Mar 3, 2024
Sunday morning read: The Shift from Models to Compound AI Systems from the Berkeley AI Research blog ☕️ A compound AI system is a system that contains multiple interacting components, such as calls to models, retrievers, or external tools. Why state-of-the-art results come
51K
Erika Shorten
@eshorten300
Aug 26, 2024
Agents creating Agents (It's getting meta) ♻️ Last week, Shengran Hu (@shengranhu), Cong Lu (@cong_ml), and Jeff Clune (@jeffclune) introduced “Automated Design of Agentic Systems” (ADAS). ADAS uses Meta Agent Search to discover agents such as the Multi-Step Peer Review Agent,
67K
Erika Shorten
@eshorten300
Jan 11, 2024
Using GPT-4 to rerank search results 🥸 I want to share two nuggets from my initial experiment: 1. Adding an example of what the output should look like helps the model follow the instructions. 2. Manually tweaking the role description to improve the ranking quality wasn't as
60K
Erika Shorten
@eshorten300
Jul 21, 2024
Agentic Reasoning and Acting is the missing piece in your RAG applications 🤖 RAG systems are usually hard-coded pipelines with just retrieval and generation, whereas Agents utilize LLMs for decision-making. The ReAct framework by @ShunyuYao12 et al. further changes the Agent
29K
Erika Shorten
@eshorten300
Mar 6, 2024
RAG with Persona 🎭 If you ask the same question to multiple people, there is a strong chance each person will have a different response. Here’s a new demo on RAG with Persona using DSPy, @cohere, and @weaviate_io! When building or using chatbots, it’s important to get a
00:00
66K
Erika Shorten
@eshorten300
Feb 28, 2024
On Monday, I released my demo building a 4-layer DSPy program to convert questions into blog posts. On the same day, @EchoShao8899, @lateinteraction, and other researchers from Stanford released STORM 🌩️ STORM is an 8-layer system that can turn topics into articles with web
Erika Shorten
@eshorten300
Feb 26, 2024
Replying to @eshorten300
Video: youtu.be/ickqCzFxWj0 Demo: github.com/weaviate/recip…
50K
Erika Shorten
@eshorten300
Feb 26, 2024
DSPy enables you to build the next generation of LLM applications 🤯 DSPy is a framework for programming language models. You build these programs by defining the components, piecing them together, and optimizing the program! I'm excited to share my new video on a hands-on demo
51K
Erika Shorten
@eshorten300
Jun 26, 2024
LLM + Memory + Planning + Tools = Agents 🤖 Last month, Job and I discussed how generative AI is shifting how companies offer customer support. How can we add more layers to our RAG apps to make it more agentic? LLM: Large language model alone Memory: Short-term and long-term
00:00
32K