BloombergGPT: A Large Language Model for Finance (Sebastian Gehrmann, PhD)

Описание: Synthetic Intelligence Forum is excited to convene a session about "BloombergGPT: A Large Language Model for Finance" with Sebastian Gehrmann, PhD (Head of NLP in the CTO office, Bloomberg).

Join us for an exploration into the fascinating intersection of financial technology and generative models. It's a field that's rich with potential, where applications span from sentiment analysis to named entity recognition, and even to answering complex questions in a conversational format.

Topic: Despite the success of large language models (LLMs) in many tasks, only a few LLMs have been trained specifically for the finance sector.

In this video, Dr. Gehrmann presents a thorough overview of BloombergGPT - a 50 billion parameter LLM that has been trained specifically on a diverse range of financial data.

The Bloomberg team has developed amongst the the most sophisticated domain-specific generative models to date using a training dataset comprising over 363 billion tokens derived from Bloomberg's vast data troves, further boosted by another 345 billion tokens from general-purpose datasets.

In this video, Dr. Gehrmann sheds light on the architectural and design choices they made for modeling, the processes they used during training, and their unique evaluation methodology.

But that's not all - Dr. Gehrmann also describes how they put BloombergGPT to the test on standard LLM benchmarks, open financial benchmarks, and a set of bespoke internal benchmarks crafted to reflect their usage intentions. They found that their mixed-dataset training method produced a model that not only outperformed others on financial tasks, but did so without falling behind on general LLM tasks.

So get ready for a comprehensive introduction into the future of financial technology and generative models.

Biography: Dr. Sebastian Gehrmann is the Head of NLP in the CTO office at Bloomberg, where he supports the development of language technology across the company. Formerly, he was a researcher at Google and holds a Ph.D. in Computer Science from Harvard University. His research interests include natural language generation, model evaluation, and interpretability. He particularly likes working on large multi-disciplinary collaborations, for example the GEM benchmark.

Profiles of the host and presenter:
• Vik Pant, PhD -   / vikpant  
• Sebastian Gehrmann, PhD -   / sebastiangehrmann  

Web profiles of Sebastian Gehrmann, PhD:
• GitHub - https://sebastiangehrmann.github.io/
• Google Scholar - https://scholar.google.com/citations?...

Join Synthetic Intelligence Forum online:
• Website - http://www.synthint.org
• LinkedIn (Page) -   / synthint  
• LinkedIn (Group) -   / 12092618  
• YouTube -    / syntheticintelligenceforum  

Special Thanks to our Partner:
• ET Business Services

Special Thanks to our Partner:
• ET Business Services