Program

Time (PST) Event
9:00 AM - 9:05 AM Welcome: Introduce organizers, keynote speakers, workshop objectives, and schedule (Dimitris Spathis)
9:05 AM - 9:40 AM Opening keynote: Ahmad Beirami (Google Research, New York, USA)
Title: Language Model Alignment: Theory & Practice
9:40 AM - 10:30 AM Paper Session #1
  • Towards Generalized Brain Decoding of visual stimuli: Cross-Subject and Dataset Perspectives with a simple alignment technique
  • LaFFi: Leveraging Hybrid Natural Language Feedback for Fine-tuning Language Models
  • Evaluating Fairness in Self-supervised and Supervised Models for Sequential Data
10:30 AM - 11:00 AM Break: Light refreshments available near session rooms
11:00 AM - 11:35 AM Keynote: Neil Zeghidour (KyutAI, France)
Title: Audio Language Models
11:35 AM - 12:30 PM Paper Session #2
  • Advancing Ante Hoc Explainable Models through Generative Adversarial Networks
  • H2G2-Net: A Hierarchical Heterogeneous Graph Generative Network Framework for Discovery of Multi-Modal Physiological Responses
  • MFA-Net: Multi-scale feature fusion attention network for liver tumor segmentation
  • Representation Learning for Wearable-Based Applications in the Case of Missing Data
12:30 PM - 2:00 PM Lunch Break
2:00 PM - 2:35 PM Keynote: Prof. Marinka Zitnik (Harvard University, USA)
Title: Towards universal models for time series
2:35 PM - 3:30 PM Paper Session #3
  • Improving Activation Steering in Language Models with Mean-Centring
  • Learning Human-like Representations to Enable Learning Human Values
  • Semi-Supervised Graph Representation Learning with Human-centric Explanation for Predicting Fatty Liver Disease
  • Balancing Continual Learning and Fine-tuning for Human Activity Recognition
3:30 PM - 3:50 PM Break: Light refreshments available near session rooms
3:50 PM - 4:25 PM Keynote: Prof. Jimeng Sun (University of Illinois Urbana-Champaign, USA)
Title: Generative AI for Clinical Trial Development
4:25 PM - 5:00 PM Keynote: Ishan Misra (Facebook AI Research/Meta, New York, USA)
Title: Generative Models for Multimodal Learning

Paper presentations are limited to a maximum of 10 minutes, followed by a 3-minute Q&A session. Attendees and authors are encouraged to gather during breaks to discuss their posters.