Eeshan Gunesh Dhekane
@EeshanDhekane
Machine Learning Researcher Engineer at MLR, Apple (Cambridge, UK) | Interested in Artificial Intelligence, Mathematics, Philosophy, and Astronomy
Happy to share that Compositional Risk Minimization has been accepted at #ICML2025 📌Extensive theoretical analysis along with a practical approach for extrapolating classifiers to novel compositions! 📜 arxiv.org/abs/2410.06303
Here is an RL perspective on understanding LLMs for decision making. Are LLMs best used as: policies / rewards / transition functions ? How do you fine-tune them ? Can LLMs explore / exploit ? 🧵 Join us down this rabbit hole... (ICLR 2025 paper, done at ML Research)
Excited to share "How to Scale Your EMA" arxiv.org/abs/2307.13813 where we look at model exponential moving averages. Q: My model is worse when I train with large batch sizes. I'm scaling my learning rate properly, what's happening? 🤔 A: You should scale your EMA momentum!💡