Natural Language Processing at Yale
Step-level Optimization for Efficient Computer-use Agents
RbtAct: Rebuttal as Supervision for Actionable Review Feedback Generation