Keep on to blur preview images; turn off to show them clearly
I should note that what we did with OlmoRL is not particularly novel; we basically did a bunch of experiments to determine what we believe to be an optimal combination of the existing literature.
modeling language at @allen_ai
Excited to see Olmo 3 up on OpenRouter: https://t.co/fz7xGfxMCs Please use it and let us know your experience!
modeling language at @allen_ai
I am gracing my colleagues at Ai2 with the gift of my being exceptionally pedantic about what, exactly, RL is
modeling language at @allen_ai
One of the biggest lessons I learned working on Olmo 3 was that it’s absolutely critical to run test scripts exactly matching your full runs before running the main one. We ran smoke tests that were close but not identical, and it caused a loooooot of pain.
modeling language at @allen_ai
Listening to this podcast reminds of the founder of a popular AI company who gave me an offer with a short exercise window when I pushed back, he explained that short exercise windows were good for the company as they helped avoid “equity debt” I did not take that offer
modeling language at @allen_ai
RT @leavittron: This is not a model release. This is a comprehensive foundation model pipeline + methodology release that includes the outp…
modeling language at @allen_ai