When the Same Prompt Gives Different Answers (and the Same Answer Isn't Always Right): A Taxonomy of Failure Mechanisms, Testing, and Mitigation Strategies for LLM-Assisted Evidence Synthesis Read-and-Evaluate Tasks
Abstract
LLM read-and-evaluate tasks include study screening, data extraction, risk-of-bias assessment, and critical appraisal in the evidence synthesis field. These task differ in the amount of LLM interpretation they require. Some tasks involve retrieving relatively simple information, such as an author name, sample size, or publication year. Others require the LLM to read source material, interpret it against predefined criteria, and make a judgement. Researchers should assess whether LLM judgements are accurate, supported by appropriate source evidence, and reproducible under specified conditions. A correct judgement does not necessarily indicate that the underlying evaluation process is reliable. We developed a taxonomy for identifying mechanisms that may compromise these properties and for informing the selection and reporting of appropriate evaluation methods. The taxonomy outlines potential ways that LLM judgments from read-and-evaluate tasks can fail, called TRAILS (Taxonomy of Reliability And Instability in LLM-assisted evidence Synthesis). The TRAILS taxonomy addresses two complementary questions. First, what mechanisms can cause the same read-and-evaluate task to produce different judgements across repeated LLM runs? Second, what mechanisms can systematically affect the judgement or required output in a way that is not apparent from the judgement itself? This study presents 38 mechanisms and identifies the deployment settings in which each can operate, approaches for testing them, and mitigation strategies. We also show why repeatability or accuracy does not rule out a hidden judgement failure and why human evaluation should assess whether a judgement is grounded in the supplied evidence, independent of inappropriate external influence, and produced through a process that can be examined and verified.
Related
- What Should We Measure Next? Finding Identification Strategies by Refining Mechanisms
- Structural taxonomy and complex irreducible representations of non-abelian groups of order p^4
- Mathematical Programming in Machine Learning and Artificial Intelligence: A Unified Taxonomy of Models and Applications
- Operational Regimes in Non-Convex Optimization: A Multiplier-Based Taxonomy
- Scientific Data Analysis for Class-Informatics in Computational Taxonomy
BibTeX
@misc{lunny2026when,
title = {{When the Same Prompt Gives Different Answers (and the Same Answer Isn't Always Right): A Taxonomy of Failure Mechanisms, Testing, and Mitigation Strategies for LLM-Assisted Evidence Synthesis Read-and-Evaluate Tasks}},
author = {Carole Lunny and Ian Goodman and Stefano Berrone and Connor TA Brenna and Emma K Reid and Ivan Perez-Neri and Shailesh Kolekar and Nityanand Jain and Ashwin Kamath and Lucas HCC Santos and Yulia Otmakhova and Heba Hussein and Diana M Cespedes and Gioia Dahdal and Salmaan Kanji},
howpublished = {arXiv},
year = {2026},
month = oct,
doi = {10.48550/arxiv.2610.39070},
eprint = {2610.39070},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2610.39070},
note = {CC BY-NC-ND 4.0},
}
RIS
TY - UNPB
TI - When the Same Prompt Gives Different Answers (and the Same Answer Isn't Always Right): A Taxonomy of Failure Mechanisms, Testing, and Mitigation Strategies for LLM-Assisted Evidence Synthesis Read-and-Evaluate Tasks
AU - Carole Lunny
AU - Ian Goodman
AU - Stefano Berrone
AU - Connor TA Brenna
AU - Emma K Reid
AU - Ivan Perez-Neri
AU - Shailesh Kolekar
AU - Nityanand Jain
AU - Ashwin Kamath
AU - Lucas HCC Santos
AU - Yulia Otmakhova
AU - Heba Hussein
AU - Diana M Cespedes
AU - Gioia Dahdal
AU - Salmaan Kanji
PY - 2026
DA - 2026/10/29
PB - arXiv
DO - 10.48550/arxiv.2610.39070
UR - https://arxiv.org/abs/2610.39070
AB - LLM read-and-evaluate tasks include study screening, data extraction, risk-of-bias assessment, and critical appraisal in the evidence synthesis field. These task differ in the amount of LLM interpretation they require. Some tasks involve retrieving relatively simple information, such as an author name, sample size, or publication year. Others require the LLM to read source material, interpret it against predefined criteria, and make a judgement. Researchers should assess whether LLM judgements are accurate, supported by appropriate source evidence, and reproducible under specified conditions. A correct judgement does not necessarily indicate that the underlying evaluation process is reliable. We developed a taxonomy for identifying mechanisms that may compromise these properties and for informing the selection and reporting of appropriate evaluation methods. The taxonomy outlines potential ways that LLM judgments from read-and-evaluate tasks can fail, called TRAILS (Taxonomy of Reliability And Instability in LLM-assisted evidence Synthesis). The TRAILS taxonomy addresses two complementary questions. First, what mechanisms can cause the same read-and-evaluate task to produce different judgements across repeated LLM runs? Second, what mechanisms can systematically affect the judgement or required output in a way that is not apparent from the judgement itself? This study presents 38 mechanisms and identifies the deployment settings in which each can operate, approaches for testing them, and mitigation strategies. We also show why repeatability or accuracy does not rule out a hidden judgement failure and why human evaluation should assess whether a judgement is grounded in the supplied evidence, independent of inappropriate external influence, and produced through a process that can be examined and verified.
ER -