?? Thrilled to announce that Collide Technology has achieved the #2 ranking worldwide on the Spider 2.0-snow dataset - one of the most challenging benchmarks for Natural Language to SQL systems! Our team took on the hardest variant of the Spider 2.0 challenge, which was developed by the Language Information and Learning at Yale University (LILY) lab run by Dragomir Radev to push the boundaries of what's possible in database interaction. While common LLMs like GPT-4 achieve high accuracy on the original Spider dataset, they struggle significantly with Spider 2.0's enterprise-level complexity - solving only about 10% of tasks. Why does this matter? Spider 2.0-snow tests an AI system's ability to translate natural language questions into complex SQL queries across huge databases (3000+ columns) using the Snowflake dialect. This isn't just about academic benchmarks - it's about making enterprise data accessible to everyone in your organization, regardless of their SQL expertise. Our Agent-Based AI system's performance on this challenge demonstrates its potential to transform how people interact with their data. Imagine empowering your entire team to get precise answers from your databases just by asking questions in plain English! Excited to be at the forefront of making enterprise data more accessible and actionable. The future of human-data interaction is conversational, and we're helping build it! #ArtificialIntelligence #DataScience #Innovation #NLP #EnterpriseAI #SQL
关于我们
- 网站
-
www.collidetech.com
Collide Technology的外部链接
- 所属行业
- 科技、信息和网络
- 规模
- 1 人
- 类型
- 自有
Collide Technology员工
动态
-
?? Operations and Coffee Office Hours with Collide Technology was buzzing with interactions this week! ?? We had an awesome discussion on engaging different parts of your operations ??, and even more exciting was the preview of our Collide Digital Factory Simulation! ????? The feedback we received was invaluable, and we also showcased some of our scheduling tools. ?? ?? Next week, we’ll dive into how to handle knowledge throughout your organization ??—you won’t want to miss it! Join us for more insights, coffee, and innovation! ?? #Operations #Technology #DigitalFactory #CollideTech #Innovation #KnowledgeManagement
-
Don't forget, tomorrow! Collide Technology is hosting Operations and Coffee Virtual Office Hour. See you all there! #Manufacturing #AI #SiliconSwamp #OperationsResearch #OfficeHours
If you're interested in some of the work we're doing over at Collide, and want to ask about AI and optimization on the operations field come join us for our new weekly office hours!
此处无法显示此内容
在领英 APP 中访问此内容等
-
?? Unlocking Insights: Context-Aware Models at Collide Technology! ?? Did you know that at Collide Technology, we've been quietly enhacing data interactions for some time now? ?? Our secret sauce? Context-aware models powered by customer and real-world data from RAG pipelines! ?? What exactly are context-aware models, you ask? Simply put, they're the key to bridging the gap between your data and Language Model (LLM) interactions. ?? By leveraging diverse datasets, we've been able to develop models that understand the nuances of your data, offering insights that are not just relevant but actionable. At Collide Technology, we're not just about breakthrough announcements; but we hope to have some on the way soon! ?? Interested in learning more about how our context-aware models can transform your data interactions? Let's chat! Reach out to us today and let's explore the possibilities together. #DataScience #ContextAwareModels #DataIntegration #CollideTechnology #RAG #AI
-
-
Super cool to hear about the experience of an open-source developer who has had such a large impact!
Investment: 9$ + some spare time. Profit: 25M ?? ↓ + 6-figure $ earnings in 3 yrs. Here is the trick and a template you can copy. I must be Joking, no ? As of Apr 2024 there has been 15.1M downloads in ?? on Parrot. Really ? In a post ChatGPT world ? were paraphrasing isn’t a big deal. ?????? ? Researchers from Amazon, CMU, Stanford, MIT, Oxford have cited Parrot as a Data Augmenting Framework across top-tier conferences / journals Like NeurIPS and ACL in 21 papers (that we know of). Utility eats Novelty for breakfast. ?????????? ? When building OSS libraries or generating ideas people tend to overload on Novelty. You don't have to always do something no one else has done. Shift your focus on to things that will help run and reproduce researches which are otherwise hard. I purely focussed on the utility of a data augmenting framework because I wanted it myself. Anserini by Jimmy Lin's Casteroni is a great example, they help reproduce IR researches easily because they are IR researcher themselves. Same is the case with FlashRank, it’s just useful to have something lite and fast. This is a theme and template you can replicate. ?????? ???? ???????? ?????????????????? ? Parrot is now part of the data augmentation research literature enjoying the network effects. People keep citing it in circles. Something similar happened to Gramformer as well. All Model downloads, GitHub stars across socials gives you eyeballs which filters into high quality consulting and advisory $ opportunities. Top 5 of the 21 papers citing Parrot: 1?? EHRSQL: A Practical Text-to-SQL Benchmark for Electronic Health Records - https://lnkd.in/ev9FjxMN (NeurIPS 2022) 2?? Targeted Data Generation: Finding and Fixing Model Weaknesses -?https://lnkd.in/e76Je5r6 (ACL 2023) 3?? OxfordTVG-HIC: Can Machine Make Humorous Captions from Image https://lnkd.in/e6YfD3mg (CVF 2023) 4?? METRIC LEARNING FOR DETECTION OF LARGE LANGUAGE MODEL GENERATED TEXTS https://lnkd.in/eJp47jdZ (ICLR 2024 - Under review?) 5?? EXPLORE, ESTABLISH, EXPLOIT: RED-TEAMING LANGUAGE MODELS FROM SCRATCH https://lnkd.in/er6fPsn6,? (ICLR 2024 - Under review?) Will be happy to share the full list of 21 papers citing. Trivia: This is the first in the history were young doctors (PhDs) are citing an old patient’s work :)