arXiv cs.AI· Yueran Tian, Huizhi Liang·· 12 小时前AI 评分29
NoveltyScope：面向科学论文证据化新颖性评估的多智能体研究助手
NoveltyScope: A Multi-Agent Research Assistant for Evidence-Grounded Novelty Assessment of Scientific Papers
AI 导读
NoveltyScope 是一个将科学论文与检索到的先前工作对比、输出结构化报告和五级新颖性评分的原型系统，采用 Prosecutor、Defender、Judge、Critic 四个评分角色，Judge 评分最多修订一次。
正文 · 原文
Abstract:NoveltyScope is a prototype that compares a scientific paper with retrieved prior work and produces a structured report and a five-level novelty score. The system extracts the core task and at most two author-claimed contributions, then searches a closed local corpus using BM25 and dense retrieval with reciprocal rank fusion. Contribution comparisons request paired passages for potential refutations, and quotations are checked against cached source text. Four scoring roles (Prosecutor, Defender, Judge, and Critic) use the resulting evidence, with at most one revision of the Judge's score. Evaluation uses 50 class-balanced RINoBench samples whose idea-level labels are retained for full-paper input. Mean absolute error is 1.48 and 1.26 for single-LLM and multi-agent scoring with Gemini, respectively, and 1.22 and 1.16 with Qwen3. The vector-similarity baseline has an error of 1.68. A HiCLIP case study illustrates the contribution comparisons and paired evidence passages. Overall, the results demonstrate the feasibility of evidence-grounded novelty assessment and show that the multi-agent configurations achieve lower scoring errors than their single-LLM counterparts across both evaluated models.
| Comments: | 13 pages, 1 figure |
| Subjects: | Information Retrieval (cs.IR); Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2610.22703 [cs.IR] |
| (or arXiv:2610.22703v1 [cs.IR] for this version) | |
| https://doi.org/10.48550/arXiv.2610.22703 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Yueran Tian [view email]
[v1]
Thu, 15 Oct 2026 22:09:25 UTC (1,236 KB)
来源：arXiv cs.AI · arxiv.org