arXiv cs.CL· Melika Mobini, Andres Algaba, Vincent Ginis·· 17 小时前AI 评分31
提高推理强度可提升 GPT-5.5 参考文献可验证性并改变推理摘要
Higher Reasoning Effort Improves Reference Verifiability and Changes Reasoning Summaries in GPT-5.5
AI 导读
在 1000 篇论文的配对实验中，GPT-5.5 将推理强度从低调至高后，经解析器验证的论文级参考文献平均通过率从 78.3% 升至 84.8%，推理摘要也变得更长。四项预设特征在高强度下得分均更高，而 GPT-5.5 自行发现的附加特征变化不一，部分在两种条件下出现频率相同。在各自推理强度内，两类特征的得分升高都未带来验证率的持续提升。
正文 · 原文
Abstract:Large language models are increasingly being used in different stages of scientific knowledge production, including the suggestion of references. These models can operate at different levels of reasoning effort and provide summaries of their reasoning processes. This raises two questions: Does increasing reasoning effort improve the verifiability of generated references, and do the resulting reasoning summaries reveal how the model's behavior changes with greater effort? We examined summaries from GPT-5.5 as it generated scholarly references for 1,000 papers under paired low- and high-effort settings. Generated references were matched to external bibliographic records, and reasoning summaries were scored for four predefined features and explored for additional features discovered by GPT-5.5. High effort increased the mean paper-level resolver-verified reference rate from 78.3% to 84.8%, and produced longer reasoning summaries. All four predefined features received higher scores under high effort. The discovered features changed less uniformly: some appeared equally often in both conditions, whereas others were more common under high effort. Within each effort condition, neither feature set showed a consistent increase in verification rates with higher scores.
| Comments: | 4 main figures. Supplementary material included |
| Subjects: | Computation and Language (cs.CL) |
| Cite as: | arXiv:2610.34212 [cs.CL] |
| (or arXiv:2610.34212v1 [cs.CL] for this version) | |
| https://doi.org/10.48550/arXiv.2610.34212 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Melika Mobini [view email]
[v1]
Sun, 25 Oct 2026 10:57:05 UTC (2,065 KB)
来源：arXiv cs.CL · arxiv.org