{"dcterms:modified":"2026-05-14","dcterms:creator":"Harvard Dataverse","@type":"ore:ResourceMap","schema:additionalType":"Dataverse OREMap Format v1.0.2","dvcore:generatedBy":{"@type":"schema:SoftwareApplication","schema:name":"Dataverse","schema:version":"6.10.1 build iqss-3","schema:url":"https://github.com/iqss/dataverse"},"@id":"https://dataverse.harvard.edu/api/datasets/export?exporter=OAI_ORE&persistentId=https://doi.org/10.7910/DVN/OV2WAM","ore:describes":{"citation:keyword":[{"citation:keywordValue":"Large Language Models"},{"citation:keywordValue":"Prompt Injection","citation:keywordVocabulary":"Generative AI"},{"citation:keywordValue":"Education Sciences"},{"citation:keywordValue":"Education Evaluation"},{"citation:keywordValue":"Trustworthy AI"},{"citation:keywordValue":"Human-in-the-Loop AI"}],"citation:dsDescription":{"citation:dsDescriptionValue":"The dataset contains the output of experiments on a research project on \nVulnerability of LLMs in Educational Assessment.\n\nThe Dataset contains:\n-the students assignments data in normal form and the injected form\n-the output produced by the experimented LLMs: ChatGPT, Gemini, DeepSeek, Grok, Perplexity and Copilot for the experiments evaluation the assignments, as a single document and collectively as a group of documents, denominated:\n \n-User Legitimate LLMs Prompts\n-Normal (no injection) providing the reference base evaluation\n -Prompt Injection Pass, one  type of injection experiments, called Fail-To-Top,  to move an assignment evailuated FAIL by reference base evaluation to PASS, i.e. above 35% of total points.\n -Prompt Injection to Top25 , a type of injection experiments  to move to top 25% an assignment with lowe reference base evaluation . This latter type of experiment come in 3 versions, Fail-To-Top, Sat-To-Top, Good-To-Top where assignment with reference base evaluation respectively: Fail (below 35%), Satisfactory (greater than 25% and belo 50%) and Good (above 50% and below 75%) are considered for injection.\n\nThe name of the folders and output results files are accordingly self-explanatory .","citation:dsDescriptionDate":"2025-08-31"},"citation:datasetContact":[{"citation:datasetContactName":"Milani, Alfredo","citation:datasetContactAffiliation":"Link Campus University, Rome, Italy","citation:datasetContactEmail":"a.milani@unilink.it"},{"citation:datasetContactName":"Valentina Franzoni","citation:datasetContactAffiliation":"University of Perugia, Italy","citation:datasetContactEmail":"valentina.franzoni@unipg.it"},{"citation:datasetContactName":"Florindi Emanuele","citation:datasetContactAffiliation":"University of Modena-Reggio Emilia","citation:datasetContactEmail":"emanuele.florindi@unimore.it"}],"author":{"citation:authorName":"Milani, Alfredo","citation:authorAffiliation":{"scheme":"http://www.grid.ac/ontology/","termName":"Link Campus University","@type":"https://schema.org/Organization","@id":"https://ror.org/035mh1293"},"authorIdentifierScheme":"ORCID","authorIdentifier":{"personName":"MILANI, Alfredo","@id":"https://orcid.org/0000-0003-4534-1805","scheme":"ORCID","@type":"https://schema.org/Person"}},"publication":{"publicationCitation":"\"When AI is Fooled: Hidden Risks in LLM-assisted Grading\"\nAuthors:\nAlfredo Milani, Valentina Franzoni, Emanuele Florindi, Assel Omarbekova, Gulmira\nBekmanova, Banu Yergesh\nin\nEducation Sciences, ISSN 2227-7102","publicationIDType":"issn","publicationIDNumber":"2227-7102","publicationRelationType":"IsSupplementTo"},"citation:depositor":"Milani, Alfredo","title":"Vulnerability of LLMs in Educational Assessment","dateOfDeposit":"2025-09-12","subject":["Computer and Information Science","Social Sciences"],"@id":"https://doi.org/10.7910/DVN/OV2WAM","@type":["ore:Aggregation","schema:Dataset"],"schema:version":"1.0","schema:name":"Vulnerability of LLMs in Educational Assessment","schema:dateModified":"Fri Sep 12 06:47:47 EDT 2025","schema:datePublished":"2025-09-12","schema:creativeWorkStatus":"RELEASED","schema:license":"http://creativecommons.org/publicdomain/zero/1.0","dvcore:fileTermsOfAccess":{"dvcore:fileRequestAccess":true},"schema:includedInDataCatalog":"Harvard Dataverse","schema:isPartOf":{"schema:name":"Harvard Dataverse","@id":"https://dataverse.harvard.edu/dataverse/harvard","schema:description":"<span><span><span><h3>Share, archive, and get credit for your data. Find and cite data across all research fields.</h3></span></span></span>"},"ore:aggregates":[{"schema:description":"The dataset contains the output of experiments on a research project on \nVulnerability of LLMs in Educational Assessment.\n\nThe Dataset contains:\n-the students assignments data in normal form and the injected form\n-the output produced by the experimented LLMs: ChatGPT, Gemini, DeepSeek, Grok, Perplexity and Copilot for the experiments evaluation the assignments, as a single document and collectively as a group of documents, denominated:\n  \n-Normal (no injection) providing the reference base evaluation\n -Prompt Injection Pass, one  type of injection experiments, called Fail-To-Top,  to move an assignment evailuated FAIL by reference base evaluation to PASS, i.e. above 35% of total points.\n -Prompt Injection to Top25 , a type of injection experiments  to move to top 25% an assignment with lowe reference base evaluation . This latter type of experiment come in 3 versions, Fail-To-Top, Sat-To-Top, Good-To-Top where assignment with reference base evaluation respectively: Fail (below 35%), Satisfactory (greater than 25% and belo 50%) and Good (above 50% and below 75%) are considered for injection.\n\nThe name of the folders and output results files are accordingly self-explanatory .","schema:name":"Normal_and_Injected_Assignment_Experiments.zip","dvcore:restricted":false,"schema:version":1,"dvcore:datasetVersionId":503127,"@id":"https://dataverse.harvard.edu/file.xhtml?fileId=12068843","schema:sameAs":"https://dataverse.harvard.edu/api/access/datafile/12068843","@type":"ore:AggregatedResource","schema:fileFormat":"application/zip","dvcore:filesize":4804924,"dvcore:storageIdentifier":"s3://dvn-cloud:1993d867500-4f795b148d7e","dvcore:rootDataFileId":-1,"dvcore:checksum":{"@type":"http://www.w3.org/2001/04/xmldsig-more#md5","@value":"d6580deb1f5fd647a0b3f3ccbb31fbda"}}],"schema:hasPart":["https://dataverse.harvard.edu/file.xhtml?fileId=12068843"]},"@context":{"author":"http://purl.org/dc/terms/creator","authorIdentifier":"http://purl.org/spar/datacite/AgentIdentifier","authorIdentifierScheme":"http://purl.org/spar/datacite/AgentIdentifierScheme","citation":"https://dataverse.org/schema/citation/","content":"@value","dateOfDeposit":"http://purl.org/dc/terms/dateSubmitted","dcterms":"http://purl.org/dc/terms/","dvcore":"https://dataverse.org/schema/core#","lang":"@language","ore":"http://www.openarchives.org/ore/terms/","personName":"https://schema.org/name","publication":"http://purl.org/dc/terms/isReferencedBy","publicationCitation":"http://purl.org/dc/terms/bibliographicCitation","publicationIDNumber":"http://purl.org/spar/datacite/ResourceIdentifier","publicationIDType":"http://purl.org/spar/datacite/ResourceIdentifierScheme","publicationRelationType":"http://datacite.org/schema/kernel-4/simpleTypes#relationType","schema":"http://schema.org/","scheme":"http://www.w3.org/2004/02/skos/core#inScheme","subject":"http://purl.org/dc/terms/subject","termName":"https://schema.org/name","title":"http://purl.org/dc/terms/title"}}