Coursera Flash Sale
40% Off Coursera Plus for 3 Months!
Grab it
Watch a technical seminar presentation exploring Llemma, a groundbreaking open language model designed specifically for mathematics. Learn about the development of 7 billion and 34 billion parameter models initialized with Code Llama weights and trained on the extensive Proof-Pile II dataset containing 55 billion tokens of mathematical content, code, and scientific papers. Discover how these models demonstrate enhanced mathematical capabilities, outperforming the Minerva model suite on comparable parameters, while showing proficiency in tool use and formal theorem proving without additional fine-tuning. Understand the significance of openly releasing all components, including models, dataset, and replication code, enabling further research and tool development at the intersection of generative AI and mathematics. Presented by Sean Welleck from CMU's Language Technologies Institute at the Harvard CMSA New Technologies in Mathematics Seminar.
