Loading activity…
@micahgoldblum
🤖Prof at Columbia University 🏙️. All things machine learning.🤖
AI
The author says Chronos-2 cut next-year prediction error by 37% against pooled past scores across 19 models and 12 datasets.
AI
The announcement claims large performance boosts over standard algorithms like DAPO.
AI
Its creators say the lightweight model uses the prompt, answer and target model’s ID to predict whether an answer is correct.
