Avi Trost
@atrost.bsky.social
PhD Student @UW-Madison, working on synthetic data, instruction tuning, and foundation models, @BrownUniversity '24
https://avitrost.github.io/
https://avitrost.github.io/
Check out this awesome work: well-curated data is and will continue to be key for superhuman LLM performance
Today at @iclr-conf.bsky.social, come chat with @changho.bsky.social about what types of data drive weak-to-strong generalization!
April 23, 2025 at 8:34 PM
Check out this awesome work: well-curated data is and will continue to be key for superhuman LLM performance
Reposted by Avi Trost
Tons of model weights available, but what else can we do besides prediction? 🤔 Introducing Grad-Mimic! A new data selection framework using well-trained model’s weights to find high-value samples for foundation models. Boost data curation & data efficiency!
February 9, 2025 at 9:08 PM
Tons of model weights available, but what else can we do besides prediction? 🤔 Introducing Grad-Mimic! A new data selection framework using well-trained model’s weights to find high-value samples for foundation models. Boost data curation & data efficiency!
Reposted by Avi Trost
What enables a strong model to surpass its weaker teacher?
🚀 Excited to share our ICLR 2025 paper: "Weak-to-Strong Generalization Through the Data-Centric Lens"! 🧵
🚀 Excited to share our ICLR 2025 paper: "Weak-to-Strong Generalization Through the Data-Centric Lens"! 🧵
February 5, 2025 at 6:22 PM
What enables a strong model to surpass its weaker teacher?
🚀 Excited to share our ICLR 2025 paper: "Weak-to-Strong Generalization Through the Data-Centric Lens"! 🧵
🚀 Excited to share our ICLR 2025 paper: "Weak-to-Strong Generalization Through the Data-Centric Lens"! 🧵