WHAT: TWed Paper Discussion: Danielle Villa on "DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning"
WHEN: 4p, Weds, 19 Feb 2025 (pizza et.al. 3:30p)
WHERE: Winslow 1140
VIDEO: https://youtu.be/bZf82ukgpgU
SLIDES: TBD
EVENT PAGE: https://bit.ly/4hWQ4Ts
Daniella Villa leads us in a discussion of "DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning" (DeepSeek-AI, et.al.). This is the second of our TWed Paper Discussion series.
TWed Paper Discussions do not require pre-reading the papers being discussed, but feel free to read them over ahead of time; Wednesday's paper may be found at: https://arxiv.org/abs/2501.12948
Danielle is a 3rd year PhD student under Deborah McGuinness studying how semantic technologies can be used to improve LLMs, with a focus on improving evaluation metrics using knowledge graphs. Her current work is on a process for generating counterfactuals to question answering datasets using knowledge graphs to evaluate the faithfulness of LLM-generated explanations.
Pizzas and salads will be delivered at approx. 3:30p; the talk will begin at 4p.