Large language models struggle with detailed novelty checks of AI papers
NovGauge: A Fine-Grained Benchmark for Diagnosing LLMs' Capability in Paper Novelty Assessment
Artificial Intelligence
Summary
Assessing how new or original a scientific paper is can be tricky for large language models (LLMs) that help with AI peer review. The authors created NovGauge, a detailed test set that breaks down novelty into parts like tasks, problems, and methods, to see exactly where models fail. They found many models either hallucinate or give reasons not supported by evidence. Even the best model only gets partial scores when checked carefully, showing that current AI tools are still unreliable at judging scientific novelty.
What this means in practice
- •For conference organizers: Use NovGauge to evaluate and improve AI tools that assist peer review by diagnosing LLMs' ability to assess scientific novelty correctly.
- •For academic publishers: Incorporate fine-grained novelty diagnostics from NovGauge to strengthen automated checks for originality in submitted research papers.
Authors
Guoqiang Zhang, Kexin Tan, Ming Zhang, Li Ju, Wenqing Jing, Zhonghan Yue, Jiayi Chen, Shiqiang Wu, Shaofan Liu, Yue Zhang, Yuankai Ying, Yang Shi, Tao Gui, Qi Zhang, Xuanjing Huang
Abstract
Large language models (LLMs) are increasingly used in peer review at major AI conferences, yet novelty remains a persistent weak point. Existing benchmarks assess novelty as a single holistic score, making it difficult to diagnose which dimension a model misjudges or whether its evidence is faithful. We present NovGauge, a human-anchored benchmark for fine-grained novelty assessment diagnosis. The benchmark contains 619 paper pairs and 50 multi-paper sets, drawn from two expert sources: ICLR reviewer overlap claims and survey co-citations. Instances are independently labeled along three dimensions: task, problem, and method, capturing application goals, technical challenges, and solution approaches. We propose a cascading diagnostic pipeline that verifies per-dimension correctness, evidence grounding, and logical support. Evaluation of 18 LLMs shows hallucination rates ranging from 0% to 39% across dimensions, and among non-hallucinated correct-positive judgments, over 70% cite evidence fails to logically support the stated reason. The best-performing model, GPT-5.5, achieves 43-72% Verified F1 across dimensions, while most models retain less than half of their raw F1 after faithfulness verification. These results suggest that current LLMs remain far from reliable scientific novelty assessment, particularly when correctness is conditioned on faithful evidence grounding.