Vectara

?? Milestone Reached: 2 Million Downloads!?? We are thrilled to announce that our Hughes Hallucination Detection Model (HHEM) has officially reached 2 million cumulative downloads on Hugging Face! What an exciting journey it's been—HHEM launched just last November, and only three weeks ago, we celebrated 1 million downloads. Now, with our downloads doubling, we couldn’t be more grateful to the incredible community of researchers, developers, and AI experts supporting this work. This recent acceleration is a very clear indication that the RAG AI Assistant market is maturing and moving into more production mission-critical use cases where the accuracy of the results is paramount. A massive thank you to all who are using HHEM to bring more factual consistency to AI Assistants and Agents. Here’s to even more innovation and growth ahead! Want to learn more about HHEM’s journey and how we’re continuing the fight against LLM hallucinations? Check out our latest blog post for all the details: https://bit.ly/4h1f8sy ?? #GenAI #NLP #HallucinationDetection #MachineLearning #RAGaaS

?? RAG on the Road: Dallas – December 10th ?? Join us for an exclusive deep dive into advanced Retrieval-Augmented Generation (#RAG) design patterns tailored for decision-makers, product owners, and AI developers. ?? Where: Dallas, TX ?? When: December 10th, 3:00 PM – 5:00 PM CST You’ll have the opportunity to gain: ? Strategies for enterprise-grade AI implementations ? Insights from industry experts ? Networking with top leaders in AI Proudly sponsored by @Google! ??? Spots are limited—Register now! https://bit.ly/3Op9i7k

Curious about how the real innovators are turning #GenerativeAI buzz into breakthrough solutions? ??? Listen to the Generation AI podcast, where Vectara's Ofer Mendelevitch and Sean Anderson break it down with AI leaders and dive into actionable strategies for navigating the evolving AI landscape. Follow on Spotify today ??

?? Exciting News! ?? Vectara is bringing #RAG on the Road to Dallas for an exclusive learning experience at the Google offices! ?? This is your chance to delve into the latest in Retrieval-Augmented Generation (RAG) alongside industry leaders and peers in an intimate, hands-on setting. ? Special Highlight: We’re honored to have our co-founder, Amr Awadallah, presenting at the event. With a distinguished career that includes co-founding Cloudera, Vice President at Yahoo, and Vice President at Google Cloud, Amr brings unparalleled expertise and insights into the AI landscape. Don’t miss the opportunity to learn from one of the pioneers in this space as he shares practical strategies to harness the power of RAG in your organization. ??? Date: 12/10/2024 ?? Location: Google Offices, Dallas ?? Register here: https://lnkd.in/gsnM-2GT Seats are limited, so secure yours today and be part of this transformative discussion! Let’s explore the future of AI together—see you there! ?? #AI #RAG #Innovation # #LearningOpportunity #DFW

Join us with Arize AI @ GitHub HQ to learn how to build an Agentic RAG application in literal MINUTES ??

Join Ofer Mendelevitch, our Head of Developer Relations, at the Developer Bootcamp: Mastering Multi-Agent Frameworks and Evaluation Techniques hosted by Arize AI! ?? When: December 4, 5:00 PM – 9:00 PM PST ?? Where: GitHub HQ, San Francisco Dive into a packed agenda of expert talks, hands-on workshops, and valuable networking opportunities. Don’t miss Ofer’s talk: ??? 6:40 PM – 7:00 PM: Build an Agentic RAG Application in Minutes You won't want to miss out on: ? Insights from industry leaders ? Hands-on experience with multi-agent frameworks ? Networking with fellow tech enthusiasts ? Complimentary food and drinks ?? Secure your spot now: https://bit.ly/3V91SZE #EnterpriseAI #RAGaaS #GenAI #TechMeetup

Hello Friends of Vectara! We recently hosted our first RAG on the Road in NYC and DC!?? ?? ??... Please have a review below. A few of the major highlight's were the audiences enthusiasm for: 1?? Future of RAG 2?? AI Adoption in Financial Services 3?? Common Pitfalls in DIY RAG 4?? Vectara’s vision for the future with our roadmap 5?? Build vs. Buy considerations 6?? Agentic RAG 7?? Strategies for Mitigating & Preventing Hallucinations in AI 8?? Hallucination Correction Model 9?? AI Assistants for the Enterprise If your interested in the educational material from the events please let me know and I will send over. We also have 5 more events coming to a city near you! Thx! Joe Amr Awadallah Jeff Chapman Eva Nahari Justin Hayes Tallat M. Shafaat Suleman Kazi Ofer Mendelevitch Sean Anderson Miluska Berta Shane Connelly

Is your AI tripping? Let's find out! I'm kind of obsessed with leaderboards for large language models. They're a great way to learn more about how models are performing in the real world. I just learned about the "Hallucination Leaderboard," which is computed using Vectara's?Hughes Hallucination Evaluation Model. It measures how often an LLM introduces inaccuracies when summarizing a given document.? Here's some of the current standings that might surprise you: ?? ChatGPT 4o ranks near the top of the list, with a hallucination rate of 1.5%. OpenAI's o1 mini actually beats that by just a hair, with a rate of 1.4%. The full o1 preview model clocked in at 3.3%. ??Google's Gemini 1.5 Flash 002, which powers the free version of the Gemini app, had a 3.4% hallucination rate; meanwhile, the paid version, Google Gemini 1.5 Pro 002, scored at 6.6%. ??Anthropic's Claude 3.5 Sonnet came in a bit higher than most of OpenAI's most current models, with a hallucination rate of 4.6%. ??Meta's highest ranking model was the Llama 3.1 405B Instruct, which produced a 3.9% hallucination rate when evaluated. The rate for Llama's smaller model, the 3.1 70B Instruct, was about 1% higher, at 5%. ??And the winner is...the GLM-4-9B-Chat model, developed by China-based Zhipu AI, with a hallucination rate of just 1.3%. Check out the full leaderboard at https://lnkd.in/g4FqaHVS Any of these results surprise you? How did your favorite model do? Does having this information change how you feel about the LLMs you use most often? #leaderboard #modeleval #ai

Semantic chunking vs. fixed-size chunking: Which delivers better results in Retrieval-Augmented Generation (#RAG) systems? ?? In our latest blog, Renyi Qu and Forrest Sheng Bao, engineers at Vectara, challenge the hype around semantic chunking. Through systematic evaluation across RAG tasks, they uncover surprising insights into performance, computational costs, and real-world practicality. ?? Key Takeaways: ? Fixed-size chunking often outperforms semantic chunking in real-world scenarios. ? Computational costs of semantic chunking don’t always justify the benefits. ? Focus on embedding quality for bigger gains in RAG performance. ??Check out the full blog here: https://bit.ly/3Z7f56j

Vectara recently wrapped up our first RAG on the Road events in NYC and DC, and they were a blast! ?????? Check out some of the highlights below: 1?? The Future of #RAG & #AI #Assistants for the Enterprise 2?? AI Adoption in Financial Services 3?? Common Pitfalls in DIY RAG / Build vs. Buy Considerations 4?? Vectara’s Vision and Roadmap for the Future 5??Agentic RAG 7?? Strategies for Mitigating & Preventing #AI #Hallucinations & Correction Models If you're interested in the educational materials from these events, let me know—I’d be happy to share them! We’re also thrilled to announce that we have five more events coming soon to a city near you. Stay tuned!