External resources

Material from outside Renaissance Labs that it considers relevant to its work. They are not made by Renaissance Labs; each page names the original author and links to the source.

• Validating RL environments, by Marianna13: a validation protocol and a list of validated datasets.
• Mid-training and SFT data, by Marianna13: building the mid-training mixture, extending a 9B model to 64K and 128K context, and the SFT v2 data.