Confidence-guided Refinement Reasoning for Zero-shot Question Answering
This work addresses the need for more robust and reliable reasoning in zero-shot QA, though it is incremental as it builds on existing QA models without introducing new training methods.
The paper tackles the problem of improving zero-shot question answering across text, image, and video domains by proposing C2R, a training-free framework that refines sub-questions and answers to enhance confidence scoring, resulting in consistent performance gains across diverse models and benchmarks.
We propose Confidence-guided Refinement Reasoning (C2R), a novel training-free framework applicable to question-answering (QA) tasks across text, image, and video domains. C2R strategically constructs and refines sub-questions and their answers (sub-QAs), deriving a better confidence score for the target answer. C2R first curates a subset of sub-QAs to explore diverse reasoning paths, then compares the confidence scores of the resulting answer candidates to select the most reliable final answer. Since C2R relies solely on confidence scores derived from the model itself, it can be seamlessly integrated with various existing QA models, demonstrating consistent performance improvements across diverse models and benchmarks. Furthermore, we provide essential yet underexplored insights into how leveraging sub-QAs affects model behavior, specifically analyzing the impact of both the quantity and quality of sub-QAs on achieving robust and reliable reasoning.