GCoT-Decoding: Unlocking Deep Reasoning Paths for Universal Question Answering
This addresses a bottleneck in universal question answering for AI applications, though it is incremental as it builds on existing CoT-decoding methods.
The paper tackled the limitation of existing CoT-decoding methods, which only work for problems with fixed answer sets, by proposing GCoT-decoding to extend applicability to a broader range of question-answering tasks, achieving significant improvements on free QA while maintaining strong performance on fixed QA.
Chain-of-Thought reasoning can enhance large language models, but it requires manually designed prompts to guide the model. Recently proposed CoT-decoding enables the model to generate CoT-style reasoning paths without prompts, but it is only applicable to problems with fixed answer sets. To address this limitation, we propose a general decoding strategy GCoT-decoding that extends applicability to a broader range of question-answering tasks. GCoT-decoding employs a two-stage branching method combining Fibonacci sampling and heuristic error backtracking to generate candidate decoding paths. It then splits each path into a reasoning span and an answer span to accurately compute path confidence, and finally aggregates semantically similar paths to identify a consensus answer, replacing traditional majority voting. We conduct extensive experiments on six datasets covering both fixed and free QA tasks. Our method not only maintains strong performance on fixed QA but also achieves significant improvements on free QA, demonstrating its generality.