Devin, Branch-Train-MiX, Chronos, and Multimodal insights

Here’s what caught our eye this past week.


Yi: Open Foundation Models by 01.AI

  • The authors introduce the Yi family of models, which include language and multimodal models, that achieve strong scores on popular benchmarks like MMLU and Chatbot Arena.
  • Paper.

Gemini 1.5: Unlocking multimodal understanding across millions of tokens of context

  • Google releases Gemini 1.5, which surpasses Gemini 1.0 Ultra’s performance across a broad set of benchmarks.
  • Paper

Branch-Train-MiX: Mixing Expert LLMs into a Mixture-of-Experts LLM

  • A method that trains LLMs asynchronously and then mixes them together into a Mixture-of-Experts LLM, in order to create LLMs specialized in multiple domains (coding, math).
  • Paper

MM1: Methods, Analysis & Insights from Multimodal LLM Pre-training

  • An investigation into the data and architecture choices of what makes a successful LLM.
  • Paper

Chronos: Learning the Language of Time Series

  • A new framework for understanding time series data to improve forecasting tasks.
  • Paper

Stay up to date

Interested in future weekly updates? Stay up to date by joining our Slack Community!