GCoT-Decoding: Unlocking Deep Reasoning Paths for Universal Question Answering
Guanran Luo, Wentao Qiu, Zhongquan Jian, Meihong Wang, Qingqiang Wu · 2026
Read GCoT-Decoding: Unlocking Deep Reasoning Paths for Universal Question Answering by Guanran Luo, Wentao Qiu, Zhongquan Jian, Meihong Wang, Qingqiang Wu (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2604-06794.txt.
Description
Luo and four colleagues free chain-of-thought decoding from fixed answer sets. Prompt-free CoT-decoding builds reasoning paths without hand-built prompts, but only for set-answer problems. Their GCoT-decoding branches in two stages, combining Fibonacci sampling with error backtracking to propose paths. Each path splits into reasoning and answer spans for confidence scoring, and similar paths merge into a consensus answer instead of a plain majority vote. Tests on six datasets covering fixed and free questions keep strong fixed-task scores while lifting free-task results.
Catalog record
| CREATOR | Guanran Luo, Wentao Qiu, Zhongquan Jian, Meihong Wang, Qingqiang Wu |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 54656 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2604.06794