Epsilla (YC S23)的动态

Advanced RAG Optimization with Hypothetical Questions In our previous blog post, we highlighted the issue of misalignment between question and document embedding spaces and proposed a solution to “drag” the question closer to the document embedding space using Hypothetical Document Embedding (HyDE). Today, we are introducing its “sibling” approach — Hypothetical Questions — that “drags” the documents closer to the question embedding space. See how to implement Hypothetical Questions to optimize AI Agent performance on Epsilla (YC S23) without writing a single line of code, and see the significant accuracy improvement using Epsilla’s latest agent evaluation functionality. Read more here https://lnkd.in/eB3VNfMB #AI #AIAgent #RAG #HypotheticalQuestions #Epsilla

Want to gain better customer insights from your surveys? Join our SurveyVista 2.600 Release webinar to learn about the latest AI-powered features and enhancements. Get ready to transform your data collection and feedback analysis! ?? Register now: https://lnkd.in/gJ_wTk47 ?? Date: June 27, 2024 9am PST / 12pm EST / 5pm GMT #SurveyVista #Webinar #AI #DataCollection #CustomerFeedback

?? Introducing our new dashboards, designed to enhance your LLM applications evaluation experience: 1?? Evaluate LLM Applications: Use metrics like relevance, factual accuracy, and more to measure the performance of your LLM applications. 2?? Compare Prompts: Easily compare different versions of prompts to choose the best fit for your use case. 3?? Build Your Own Experiments: Create and manage experiments effortlessly. 4?? Set Up Daily Monitoring: Keep track of your progress with daily monitoring graphs, ensuring your LLM applications are always performing at their best. Check out these dashboards here: https://lnkd.in/gaSYt8Ev #UpTrain #LLM #AI #MachineLearning #Dashboards #Productivity

?? Is Your Enterprise Search Strategy Delivering Real ROI? Large Language Models (LLMs) are transforming AI-driven enterprise search, but how do you measure their true impact? ?? From efficiency gains to cost savings, understanding the ROI of LLMs is essential for making data-driven decisions. Check out our latest blog to explore key metrics, examples, and best practices for evaluating the value of LLMs in your search strategy. ?? Read more: https://hubs.li/Q02-92J20 #EnterpriseSearch #LLMs #AI #GoSearch #DigitalTransformation #ROI

Should we select features before or after over- or under-sampling? A common question I get in my course is whether to perform feature selection before or after resampling. While there’s no one-size-fits-all answer, here are some considerations: ?? Resampling First: The main goal of resampling is to adjust the class distribution, either by oversampling the minority class or undersampling the majority class. ?? Feature Selection After: Feature selection methods often rely on the distribution of the data. To ensure that selected features truly represent the adjusted class distributions, it makes sense to perform feature selection after resampling. This way, the selection process may benefit from the newly balanced data. ?? Experiment and Decide: Research on this topic is limited, so if you have the time, experiment with both approaches to see what works best for your specific project. How do you handle feature selection and resampling in your workflows? Share your experiences in the comments! ?? For more on handling class imbalance, check out my course here. https://buff.ly/3EKLyGh #MachineLearning #DataScience #FeatureSelection #Resampling #AI #ML #DataScienceCommunity #ModelTraining

I’ve been keeping these Monitoring and Evaluation (M&E) prompts a secret… But not anymore! These 10 AI prompts have transformed how I handle M&E—and they can do the same for you. Here’s how they help: ?Automate data collection and analysis in minutes. ?Generate insightful reports with just a few clicks. ?Identify trends and patterns you might miss manually. ?Save time by eliminating repetitive M&E tasks. ?Ensure your recommendations are data-driven and actionable. If you’re not using these, you’re missing out. Philip Adu, PhD, and I are hosting a hands-on, interactive 3-6 day workshop on how to use AI for evaluation and research. We were fully booked, but we’ve just opened a few more spots! Once these are gone, that’s it—FULL is FULL! Don’t miss out! ?? Sign up here: https://lnkd.in/ez3mWiNj #ai #ArtificialIntelligence

Fergal read a good post this week by Ethan Mollick, author of Co-Intelligence, and one of the more insightful #GenAI experts. Ethan highlights a common issue with "Ask a doc anything" type interactions: "users often expect these systems to act like search engines, leading to misunderstandings. Instead, GenAI + #RAG tools can be more like forgetful analysts, sometimes producing false positives." Fergal noticed this expectation mismatch with VisibleThread's customers too. Users expect deterministic (100% accurate) results from GenAI. Fergal dives deeper into this topic in this newsletter: https://lnkd.in/eAZVdJr6 #artificialintelligence #ai

The purpose of an organization is to machine ideas – to move concepts to commerce. In large organizations, this journey is hindered by a lack of transparency of content, persistence of conversations, and cumulativeness of computation, resulting in anecdotal analytics, disjointed programs, and unscalable work. ? Listen to Dhiraj talk about how CFOs can help their organizations embrace complexity and thrive in an algorithmic world.?https://lnkd.in/dsbCy-kJ ? #DigitalTransformation?#Productivity?#AI?#WorldofAlgorithms?#CFO

The opportunites for businesses have never been greater, but so is the complexity. As such, the role of the CFO continues to morph and take greater importance. Dhiraj Rajaram, our CEO Founder, shares his thoughts on how CFOs can champion a New Path for decision making so companies can thrive in complexity.

Want better insights from your surveys? Join our SurveyVista 2.600 Release webinar, on June 27th, to learn about the latest AI-powered features. Transform your data collection and feedback analysis! #CustomerExperience #CustomerFeedback #customercentricity #voc #voiceofcustomer