Schedule
8:15 - 9:00 Registration & breakfast
9:00 - 9:40 Intro & Overview: Irina Rish & Natalia Vassilieva
9:40 - 10:30 Plenary talk: Yuhai Tu Statistical Physics of Deep-Learning: On Learning Dynamics and Generalization
10:30 - 11:00 Coffee break
Track 1: Scaling & Emergence (room A)
11 - 11:30 Paul Bogdan (USC)
11:30 - noon Hailey Schoelkopf (Eleuther AI)
Why Has Predicting Downstream Capabilities of Frontier AI Models with Scale Remained Elusive?
Track 2: Multilingual LLMs (room B)
11:00 - 11:30 Rio Yokota (Tokyo Institute of Technology)
Continual Pre-training of Open-Source Models on Japanese Text
11:30 - noon Tatiana Shavrina (Meta)
Towards Full Linguistic Diversity in Language Models
noon - 1:30 lunch break
Track 1: Scaling & Emergence (room A)
1:30 - 2:00 - Darshil Doshi (University of Maryland)
Emergence of in-context learning and skill composition
2:00 - 3:00 - CERC-AAI Open-Source Foundation Models
Irina Rish
Overview of CERC-AAI Open Source AI EffortsAlexis Roger
Robin: A Suite of Vision-Language Models
Tejas Vaidhya
Spectra : A Comprehensive Study of Ternary, Quantized, and FP16 LLMs
Track 2: Multilingual LLMs (room B)
1:30pm - 2:00pm Neha Sengupta (G42)
Comparing adaptation and training from scratch for bilingual models
2:00pm - 2:30pm Preslav Nakov (Mohamed bin Zayed University of Artificial Intelligence)
Multilinguality Challenges and Datasets for Evaluating LLMs
2:30pm - 3:00pm Yishi Xu (Cerebras Systems)
3:30 pm - 3:45pm Coffee
3:45pm - 4:45 pm Panel
4:45 pm - 5:15 pm Open discussion & interactive working session (putting together google doc with open questions & future directions)