Xinnuo Li
Pronouns: She/Her
Research Mentor(s): Lu Wang
Research Mentor School/College/Department: Computer Science and Engineering / Engineering
Program:
Authors: Xinnuo Li, Yunxiang Zhang, Lu Wang
Session: Session 4: 1:40 pm – 2:30 pm
Poster: 70
Abstract
We propose a novel Feedback-of-Feedback (FoF) mechanism to improve answer quality in question-answering systems. Our approach consists of a base generator, a critic, and a refiner. The base generator produces an initial answer, which is then assessed by the critic. The refiner incorporates the critic’s feedback to generate a revised answer. The key innovation is the FoF component, where the base generator evaluates the critic’s feedback, determining whether to agree, disagree, or request clarification. This FoF guides the critic in refining its feedback, enabling the refiner to produce higher-quality answers. We evaluate the FoF mechanism on various question-answering tasks and demonstrate improvements in answer quality compared to baselines without FoF. Our study highlights the importance of feedback quality in iterative answer refinement and opens new avenues for research on advanced feedback mechanisms in question-answering systems.
