Verizon conducted an experiment to find the optimal Top-K value for retrieval in RAG Ablation Lab.
Public source
Publisher name
Public post
🧪 RAG Ablation Lab — Day 4/10 How many documents should we retrieve? A common assumption in RAG: "If 5 documents are good, 20 must be better." Not necessarily. Too few…
Company
Verizon
- Industry
- IT Services and IT Consulting
- Location
- Basking Ridge, US
- Company size
- 10,001+ employees
About Verizon
We get you. You want more out of a career. A place to share your ideas freely — even if they’re daring or different. Where the true you can learn, grow, and thrive. You’ll find all that here. Because we empower you. We power and empower how people live, work and play by connecting them to what brings them joy. The same is true inside our walls. We do what we love — driving innovation, creativity, and impact in the world. And wherever you go, we got your back. This is a team sport. Our V Team is a community of people who anticipate, lead, and believe that listening is where learning begins. In crisis and in celebration, we come together— lifting our communities and building trust in how we show up, everywhere & always. Want in? Join the V Team Life.
See moreLatest activity
Latest activity from Verizon
27 signals
Strategy & Corporate Development
Verizon announced a major multi-billion dollar agreement with Corning Incorporated to secure over 80 million miles of fiber through 2032.
People
Verizon hired a Retail Sales Associate for North Huntingdon
Partnerships
Verizon and Corning Incorporated reached a landmark, multi-billion dollar agreement to secure 80+ million miles of high-density optical fiber and connectivity solutions from 2027 through 2032.
Discover more
Similar signals
Similar public activity from other companies.
Research & Knowledge
Oracle
Oracle ran an experiment on a chatbot for the car insurance domain in Latvian to test embedding models and Standard RAG.
Research & Knowledge
Target
Target conducted research on improving RAG results by including bi- and cross-encoders in the process and using completely synthetic training data.
Research & Knowledge
Lettria
Lettria gave AI models the same homework of writing down every fact in a document and connecting them, and tracking output validity and seconds per document
Research & Knowledge
Avanade
Avanade published a new paper titled Dual-Cache Latent Space Communication between Heterogeneous Language Models by XKV exploring an approach where Agent A and Agent B each have their own KV cache and the system translates information directly into the receiver's KV representation.
Research & Knowledge
Towards Data Science