RIGOURATE: Quantifying Scientific Exaggeration with Evidence-Aligned Claim Evaluation

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: James, Joseph, Xiao, Chenghao, Li, Yucheng, Moosavi, Nafise Sadat, Lin, Chenghua
Format: Preprint
Published: 2026
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866909987361521664
author James, Joseph
Xiao, Chenghao
Li, Yucheng
Moosavi, Nafise Sadat
Lin, Chenghua
author_facet James, Joseph
Xiao, Chenghao
Li, Yucheng
Moosavi, Nafise Sadat
Lin, Chenghua
contents Scientific rigour tends to be sidelined in favour of bold statements, leading authors to overstate claims beyond what their results support. We present RIGOURATE, a two-stage multimodal framework that retrieves supporting evidence from a paper's body and assigns each claim an overstatement score. The framework consists of a dataset of over 10K claim-evidence sets from ICLR and NeurIPS papers, annotated using eight LLMs, with overstatement scores calibrated using peer-review comments and validated through human evaluation. It employes a fine-tuned reranker for evidence retrieval and a fine-tuned model to predict overstatement scores with justification. Compared to strong baselines, RIGOURATE enables improved evidence retrieval and overstatement detection. Overall, our work operationalises evidential proportionality and supports clearer, more transparent scientific communication.
format Preprint
id arxiv_https___arxiv_org_abs_2601_04350
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle RIGOURATE: Quantifying Scientific Exaggeration with Evidence-Aligned Claim Evaluation
James, Joseph
Xiao, Chenghao
Li, Yucheng
Moosavi, Nafise Sadat
Lin, Chenghua
Computation and Language
Scientific rigour tends to be sidelined in favour of bold statements, leading authors to overstate claims beyond what their results support. We present RIGOURATE, a two-stage multimodal framework that retrieves supporting evidence from a paper's body and assigns each claim an overstatement score. The framework consists of a dataset of over 10K claim-evidence sets from ICLR and NeurIPS papers, annotated using eight LLMs, with overstatement scores calibrated using peer-review comments and validated through human evaluation. It employes a fine-tuned reranker for evidence retrieval and a fine-tuned model to predict overstatement scores with justification. Compared to strong baselines, RIGOURATE enables improved evidence retrieval and overstatement detection. Overall, our work operationalises evidential proportionality and supports clearer, more transparent scientific communication.
title RIGOURATE: Quantifying Scientific Exaggeration with Evidence-Aligned Claim Evaluation
topic Computation and Language
url https://arxiv.org/abs/2601.04350