Kempner Graduate Fellow Symposium
This all-day symposium will feature short talks from fourth-year Kempner Graduate Fellows. Topics run the breadth of research in the Kempner — this year’s talks include machine learning interpretability, human-AI interactions, biologically informed model building, and beyond!
The symposium will be followed by a catered reception, where attendees will have the opportunity to continue conversations with the presenters and other guests. All students, faculty, staff, and interested guests are welcome.
Contact kempnereducation@harvard.edu with specific event questions.
Schedule
10:30 am: Registration opens
11:00 am – 11:05 pm: Welcome Announcements
11:05 am – 12:45 pm: AI for Science, Engineering, and Behavior
- 11:05 am – 11:30 am: Usha Bhalla – “Interpreting Temporal and Geometric Structures in LLMs”
- 11:30 am – 11:55 am: Ada Fang – “Learning and Automating Biomolecular Research with AI”
- 11:55 am – 12:20 pm: Ikechukwu Uchendu – “Vision Language Models for Computer Chip Design”
- 12:20 pm – 12:45 pm: Yongsoo Ra – “Testing the role of dopamine in off-policy motor learning”
12:45 pm – 2:00 pm: Lunch Break
2:00 pm – 3:30pm: Language Models and Meaning
- 2:00 pm – 2:25 pm: Rosie Zhao – “Matching Features, Not Tokens: Energy-Based Fine-Tuning of Language Models”
- 2:25 pm – 2:50 pm: Catherine Yeh – “Telling Stories With and About AI: Visualizing and Shaping Meaning from Unstructured Data”
- 2:50 pm – 3:30 pm: Gustaf Ahdritz and Anat Kleiman – “Sub-probabilistic Modeling for Mitigating Hallucinations”
3:30 pm – 3:55 pm: Afternoon Break
3:55 pm – 5:10 pm: Representation and Exploration in AI
- 3:55 pm – 4:20 pm: Sabarish Sainathan – “How Feature Learning Responds to Correlation Blowup and Localization”
- 4:20 pm – 4:45 pm: Yuyang Zhang – “Diffusion Policies for Improved Exploration in Online Reinforcement Learning”
- 4:45 pm – 5:10 pm: Sonia Murphy – “Using Cognitive Models to Reveal Value Trade-offs in Language Models”
- 5:10 pm – 5:15 pm: Ending Announcements
5:15 pm – 6:30 pm: Reception
Please note the building is only accessible with HUID card.
