{"id":"bb54c984f590d5005339b291baa28f7bdb8a7433094f678651d7b1473ae2227f","title":"Generative AI and the Crisis of Higher Education Assessments: Examining 'AI-Proofing' Strategies for College-Level Assessments","argument":"In an arXiv abstract, authors Bonam Mingole, Davis Yadav, Hui Min Lee, S. Shyam Sundar and Amulya Yadav describe an academic-integrity challenge: students can outsource homework, take-home assignments and online examinations to large language models (LLMs) that complete them in seconds. They examine whether assessment questions designed to resist LLMs constrain model performance. The authors collected questions from authentic college coursework across disciplines, including faculty-designed questions and questions from online courses, and tested three LLMs with and without supplementary course materials. They report that the best-performing model scored approximately 83% without supplementary materials and 85% with them. They also report differences across models, answer formats and instructor strategies, and say they identified five strategies through thematic analysis. The findings point to a practical tension for instructors: questions intended to make coursework harder for LLMs may not do so equally, and performance varied with the question and model. The abstract says the authors discuss implications for assessment design, academic-integrity policy and the ethics of deploying LLMs in education. It does not name the five strategies or provide results for each, so the abstract alone cannot show which designs constrained performance or how broadly the findings apply.","topicIds":["intelligence-and-discovery"],"domain":"science","claimType":"attributed reading","limits":"The evidence provided is the abstract, which does not name the five strategies or give their individual effects. The question set includes faculty-designed questions and questions from online courses; the abstract does not describe how representative they are of assessments beyond this collection. The abstract reports approximate scores for the best-performing model but does not provide exact results for each model, answer format or strategy.","speaker":{"name":"Mingole, Bonam, Yadav, Davis, Lee, Hui Min, Sundar, S. Shyam, Yadav, Amulya","roleAtSource":"Article author"},"source":{"title":"Generative AI and the Crisis of Higher Education Assessments: Examining 'AI-Proofing' Strategies for College-Level Assessments","publisher":"arXiv cs.AI","format":"Article","url":"https://arxiv.org/abs/2611.21476","publishedAt":"2026-11-16","dateBasis":"Date basis not provided."},"evidence":[{"text":"The authors collected 75 questions from authentic college coursework spanning 17 disciplines, contributed by 37 faculty members from ten U.S. universities and one U.K. university.","excerpt":"collecting 75 questions from authentic college coursework assessments, spanning 17 disciplines, which were contributed by 37 faculty members from ten U.S. universities and one U.K. university","url":"https://arxiv.org/abs/2611.21476","relationship":"Source-stated"},{"text":"Of the questions, 61 were faculty-designed as 'LLM-proof' and 14 came from online courses.","excerpt":"Of these, 61 questions were intentionally designed by faculty members to be 'LLM-proof', while the remaining 14 questions were sourced from online courses.","url":"https://arxiv.org/abs/2611.21476","relationship":"Source-stated"},{"text":"The authors evaluated three LLMs with and without supplementary course materials, identified five faculty strategies through thematic analysis, and tested whether those strategies affected answer accuracy.","excerpt":"evaluate three state-of-the-art LLMs' ability to correctly answer these questions with and without supplementary course materials; (ii) identify five strategies, through thematic analysis","url":"https://arxiv.org/abs/2611.21476","relationship":"Source-stated"},{"text":"The best-performing model scored approximately 83% without supplementary course materials and 85% with them.","excerpt":"the best-performing model scored approximately 83% without supplementary course materials and 85% with them","url":"https://arxiv.org/abs/2611.21476","relationship":"Source-stated"},{"text":"The authors observed differences across models, answer formats and instructor strategies.","excerpt":"we observed noteworthy differences across models, answer formats, and instructor tweaking strategies","url":"https://arxiv.org/abs/2611.21476","relationship":"Source-stated"},{"text":"The abstract does not identify the five strategies or report their individual effects, so it does not establish from this evidence which strategies constrained LLM performance.","excerpt":"test whether these strategies impact LLMs' ability to answer the questions accurately","url":"https://arxiv.org/abs/2611.21476","relationship":"Escape Meridian interpretation"}],"sourceMedia":{"status":"not_found","sourceUrl":"https://arxiv.org/abs/2611.21476","checkedAt":"2026-11-17T09:36:08.760Z","items":[],"note":"No suitable media was found by the supported page extraction."},"assumptions":[],"openQuestion":{"attribution":"Open question","text":"Which of the five instructor strategies, if any, meaningfully constrained LLM performance, and how did their effects vary across models and answer formats?"},"whatWouldChangeView":{"attribution":"Change condition not recorded","text":""},"editorialComparisons":[],"live":{"title":"Generative AI and the Crisis of Higher Education Assessments: Examining 'AI-Proofing' Strategies for College-Level Assessments","summary":"In an arXiv abstract, authors Bonam Mingole, Davis Yadav, Hui Min Lee, S. Shyam Sundar and Amulya Yadav describe an academic-integrity challenge: students can outsource homework, take-home assignments and online examinations to large language models (LLMs) that complete them in seconds. They examine whether assessment questions designed to resist LLMs constrain model performance. The authors collected questions from authentic college coursework across disciplines, including faculty-designed questions and questions from online courses, and tested three LLMs with and without supplementary course materials. They report that the best-performing model scored approximately 83% without supplementary materials and 85% with them. They also report differences across models, answer formats and instructor strategies, and say they identified five strategies through thematic analysis. The findings point to a practical tension for instructors: questions intended to make coursework harder for LLMs may not do so equally, and performance varied with the question and model. The abstract says the authors discuss implications for assessment design, academic-integrity policy and the ethics of deploying LLMs in education. It does not name the five strategies or provide results for each, so the abstract alone cannot show which designs constrained performance or how broadly the findings apply.","question":"Which of the five instructor strategies, if any, meaningfully constrained LLM performance, and how did their effects vary across models and answer formats?","claims":[{"text":"The authors collected 75 questions from authentic college coursework spanning 17 disciplines, contributed by 37 faculty members from ten U.S. universities and one U.K. university.","excerpt":"collecting 75 questions from authentic college coursework assessments, spanning 17 disciplines, which were contributed by 37 faculty members from ten U.S. universities and one U.K. university","support":"source-stated"},{"text":"Of the questions, 61 were faculty-designed as 'LLM-proof' and 14 came from online courses.","excerpt":"Of these, 61 questions were intentionally designed by faculty members to be 'LLM-proof', while the remaining 14 questions were sourced from online courses.","support":"source-stated"},{"text":"The authors evaluated three LLMs with and without supplementary course materials, identified five faculty strategies through thematic analysis, and tested whether those strategies affected answer accuracy.","excerpt":"evaluate three state-of-the-art LLMs' ability to correctly answer these questions with and without supplementary course materials; (ii) identify five strategies, through thematic analysis","support":"source-stated"},{"text":"The best-performing model scored approximately 83% without supplementary course materials and 85% with them.","excerpt":"the best-performing model scored approximately 83% without supplementary course materials and 85% with them","support":"source-stated"},{"text":"The authors observed differences across models, answer formats and instructor strategies.","excerpt":"we observed noteworthy differences across models, answer formats, and instructor tweaking strategies","support":"source-stated"},{"text":"The abstract does not identify the five strategies or report their individual effects, so it does not establish from this evidence which strategies constrained LLM performance.","excerpt":"test whether these strategies impact LLMs' ability to answer the questions accurately","support":"interpretation"}],"limitations":["The evidence provided is the abstract, which does not name the five strategies or give their individual effects.","The question set includes faculty-designed questions and questions from online courses; the abstract does not describe how representative they are of assessments beyond this collection.","The abstract reports approximate scores for the best-performing model but does not provide exact results for each model, answer format or strategy."],"mediaId":"ab00ed09a09f35c1b657d28b737e86f3cd07ef2215d3aa481ba10b2a9894b2cf","correctionReason":null,"id":"bb54c984f590d5005339b291baa28f7bdb8a7433094f678651d7b1473ae2227f","publicationId":"9f44839f-ab6f-402e-8dae-91c1f540b0ad","version":1,"sourceUrl":"https://arxiv.org/abs/2611.21476","sourceTitle":"Generative AI and the Crisis of Higher Education Assessments: Examining 'AI-Proofing' Strategies for College-Level Assessments","publisher":"arXiv cs.AI","publishedAt":null,"originalPublished":{"value":"2026-11-16","precision":"day","basis":"publisher_citation_date"},"firstObservedAt":"2026-11-17T08:28:16.131Z","createdAt":"2026-11-17T08:32:02.943Z","historical":false,"entityIds":["arxiv"],"topicIds":["intelligence-and-discovery"],"media":[{"id":"ab00ed09a09f35c1b657d28b737e86f3cd07ef2215d3aa481ba10b2a9894b2cf","kind":"article","url":"https://arxiv.org/abs/2611.21476","sourceUrl":"https://arxiv.org/abs/2611.21476","label":"Read original","provenance":"source_link","verification":{"status":"verified","checkedAt":"2026-11-17T08:30:43.190Z","method":"acquired_source","contentType":"text/html"},"recommendationReason":"Recommended by Jev for understanding this source; checked by Luna."}],"sourceMedia":{"status":"not_found","sourceUrl":"https://arxiv.org/abs/2611.21476","checkedAt":"2026-11-17T09:36:08.760Z","items":[],"note":"No suitable media was found by the supported page extraction."},"editorialDecision":{"outcome":"accept","reasonCodes":["material_development"],"rubricVersion":"stream-admission-v9","mediaId":"ab00ed09a09f35c1b657d28b737e86f3cd07ef2215d3aa481ba10b2a9894b2cf","confidence":0.65,"provider":"typesafe","model":"jev-1.13.0","inputHash":"69f418c043028d0664befb7c462bc1d321ddbc08168124c33e2aca51d63fcb4f","decidedAt":"2026-11-17T08:30:43.879Z"},"provenance":{"textHash":"1832a06040f9dc866ca08f7e6878ce8277871ca6bed1ab9a16fbfeb66bf08e0d","observationId":"2f70114a057fa4eb06dc23147655a8b1ca17a7c58b67ae60a0f3af5184eb4765","retrievedAt":"2026-11-17T08:30:43.082Z","kind":"publisher-page","preparer":{"model":"gpt-6-luna","runId":"b84dc540-8dc9-4887-8140-b3b86edeb341"},"reviewer":{"model":"gpt-6-luna","runId":"0a548ed2-e789-4a30-8792-5819058d00cb"}},"audit":{"status":"passed","reviewedAt":"2026-11-17T08:32:02.943Z","model":"anthropic/claude-opus-5.5"},"sourceAttribution":{"authors":[{"name":"Mingole, Bonam","url":null,"kind":"unknown"},{"name":"Yadav, Davis","url":null,"kind":"unknown"},{"name":"Lee, Hui Min","url":null,"kind":"unknown"},{"name":"Sundar, S. Shyam","url":null,"kind":"unknown"},{"name":"Yadav, Amulya","url":null,"kind":"unknown"}],"authorRole":"article-author","hosts":[],"status":"stated","authorRoster":{"truncated":false,"omitted":0,"assertionGroupsOmitted":0},"participation":[{"actor":{"name":"Mingole, Bonam","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Mingole, Bonam","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2611.21476","locator":"head/meta[name=author|citation_author]"},"assertionId":"39fccf78cf33935498d51564d9a7328774d0812e09a36576025333416f668d18:participation:0","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-17T09:36:08.760Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"39fccf78cf33935498d51564d9a7328774d0812e09a36576025333416f668d18","sourceUrl":"https://arxiv.org/abs/2611.21476","contentSha256":"ae53fafe9106dae3059452ed64d617823f44e9d7997053c0e0c9c019ea3e07ef","textHash":"1832a06040f9dc866ca08f7e6878ce8277871ca6bed1ab9a16fbfeb66bf08e0d","observedAt":"2026-11-17T09:36:08.760Z"}},{"actor":{"name":"Yadav, Davis","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Yadav, Davis","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2611.21476","locator":"head/meta[name=author|citation_author]"},"assertionId":"39fccf78cf33935498d51564d9a7328774d0812e09a36576025333416f668d18:participation:1","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-17T09:36:08.760Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"39fccf78cf33935498d51564d9a7328774d0812e09a36576025333416f668d18","sourceUrl":"https://arxiv.org/abs/2611.21476","contentSha256":"ae53fafe9106dae3059452ed64d617823f44e9d7997053c0e0c9c019ea3e07ef","textHash":"1832a06040f9dc866ca08f7e6878ce8277871ca6bed1ab9a16fbfeb66bf08e0d","observedAt":"2026-11-17T09:36:08.760Z"}},{"actor":{"name":"Lee, Hui Min","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Lee, Hui Min","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2611.21476","locator":"head/meta[name=author|citation_author]"},"assertionId":"39fccf78cf33935498d51564d9a7328774d0812e09a36576025333416f668d18:participation:2","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-17T09:36:08.760Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"39fccf78cf33935498d51564d9a7328774d0812e09a36576025333416f668d18","sourceUrl":"https://arxiv.org/abs/2611.21476","contentSha256":"ae53fafe9106dae3059452ed64d617823f44e9d7997053c0e0c9c019ea3e07ef","textHash":"1832a06040f9dc866ca08f7e6878ce8277871ca6bed1ab9a16fbfeb66bf08e0d","observedAt":"2026-11-17T09:36:08.760Z"}},{"actor":{"name":"Sundar, S. Shyam","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Sundar, S. Shyam","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2611.21476","locator":"head/meta[name=author|citation_author]"},"assertionId":"39fccf78cf33935498d51564d9a7328774d0812e09a36576025333416f668d18:participation:3","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-17T09:36:08.760Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"39fccf78cf33935498d51564d9a7328774d0812e09a36576025333416f668d18","sourceUrl":"https://arxiv.org/abs/2611.21476","contentSha256":"ae53fafe9106dae3059452ed64d617823f44e9d7997053c0e0c9c019ea3e07ef","textHash":"1832a06040f9dc866ca08f7e6878ce8277871ca6bed1ab9a16fbfeb66bf08e0d","observedAt":"2026-11-17T09:36:08.760Z"}},{"actor":{"name":"Yadav, Amulya","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Yadav, Amulya","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2611.21476","locator":"head/meta[name=author|citation_author]"},"assertionId":"39fccf78cf33935498d51564d9a7328774d0812e09a36576025333416f668d18:participation:4","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-17T09:36:08.760Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"39fccf78cf33935498d51564d9a7328774d0812e09a36576025333416f668d18","sourceUrl":"https://arxiv.org/abs/2611.21476","contentSha256":"ae53fafe9106dae3059452ed64d617823f44e9d7997053c0e0c9c019ea3e07ef","textHash":"1832a06040f9dc866ca08f7e6878ce8277871ca6bed1ab9a16fbfeb66bf08e0d","observedAt":"2026-11-17T09:36:08.760Z"}}],"participationCoverage":{"total":5,"returned":5,"complete":true,"fullAssertions":"source-metadata-endpoint"},"distribution":null,"publisher":null,"sourceUrl":"https://arxiv.org/abs/2611.21476","snapshotId":"39fccf78cf33935498d51564d9a7328774d0812e09a36576025333416f668d18","contentSha256":"ae53fafe9106dae3059452ed64d617823f44e9d7997053c0e0c9c019ea3e07ef","textHash":"1832a06040f9dc866ca08f7e6878ce8277871ca6bed1ab9a16fbfeb66bf08e0d","observedAt":"2026-11-17T09:36:08.760Z","basis":"publisher-page-metadata","sourceLinks":43,"linksTruncated":false,"metadataUrl":"/api/pipeline/stream/bb54c984f590d5005339b291baa28f7bdb8a7433094f678651d7b1473ae2227f/source"}},"canonicalUrl":"https://escapemeridian.com/readings/generative-ai-and-the-crisis-of-higher-education-assessments-bb54c984","formats":{"html":"/readings/generative-ai-and-the-crisis-of-higher-education-assessments-bb54c984","markdown":"/readings/generative-ai-and-the-crisis-of-higher-education-assessments-bb54c984.md","json":"/readings/generative-ai-and-the-crisis-of-higher-education-assessments-bb54c984.json"},"attribution":"Headline and summary by Escape Meridian, prepared by Luna from the attributed source. A source check is not independent confirmation of the source’s claims.","maintainedGuideLinks":[],"conditions":[],"readerContext":{"version":"reader-context-v1","attribution":"Reference context selected by Escape Meridian; separate from source evidence. Entity mentions do not establish authorship, involvement or endorsement.","terms":[{"id":"term-llm","name":"Large language model","explanation":"A language model with many learned numerical settings, used to process and generate language.","referenceUrl":"https://developers.google.com/machine-learning/glossary#large-language-model"}],"entities":[{"id":"arxiv","name":"arXiv","kind":"organization","context":"arXiv is a research manuscript repository where authors share papers that may not yet have been peer reviewed.","referenceUrls":["https://arxiv.org/","https://info.arxiv.org/help/api/index.html"],"path":"/entities/arxiv","json":"/entities/arxiv.json","markdown":"/entities/arxiv.md"}]},"researchQuestions":[],"withdrawn":false,"references":{"schemaVersion":1,"target":{"type":"reading","id":"bb54c984f590d5005339b291baa28f7bdb8a7433094f678651d7b1473ae2227f","revision":"9f44839f-ab6f-402e-8dae-91c1f540b0ad","revisionKind":"publication-id","title":"Generative AI and the Crisis of Higher Education Assessments: Examining 'AI-Proofing' Strategies for College-Level Assessments","url":"https://escapemeridian.com/readings/generative-ai-and-the-crisis-of-higher-education-assessments-bb54c984"},"source":{"type":"source","id":"https://arxiv.org/abs/2611.21476","revision":"1832a06040f9dc866ca08f7e6878ce8277871ca6bed1ab9a16fbfeb66bf08e0d","revisionKind":"sha256-source-text","title":"Generative AI and the Crisis of Higher Education Assessments: Examining 'AI-Proofing' Strategies for College-Level Assessments","url":"https://arxiv.org/abs/2611.21476"},"status":"available","items":[],"next":null,"scope":"Explicit citations in public readings to this reading. Shared topics are not citations. Other record types need an explicit reference adapter."},"revisions":[{"id":"bb54c984f590d5005339b291baa28f7bdb8a7433094f678651d7b1473ae2227f","title":"Generative AI and the Crisis of Higher Education Assessments: Examining 'AI-Proofing' Strategies for College-Level Assessments","argument":"In an arXiv abstract, authors Bonam Mingole, Davis Yadav, Hui Min Lee, S. Shyam Sundar and Amulya Yadav describe an academic-integrity challenge: students can outsource homework, take-home assignments and online examinations to large language models (LLMs) that complete them in seconds. They examine whether assessment questions designed to resist LLMs constrain model performance. The authors collected questions from authentic college coursework across disciplines, including faculty-designed questions and questions from online courses, and tested three LLMs with and without supplementary course materials. They report that the best-performing model scored approximately 83% without supplementary materials and 85% with them. They also report differences across models, answer formats and instructor strategies, and say they identified five strategies through thematic analysis. The findings point to a practical tension for instructors: questions intended to make coursework harder for LLMs may not do so equally, and performance varied with the question and model. The abstract says the authors discuss implications for assessment design, academic-integrity policy and the ethics of deploying LLMs in education. It does not name the five strategies or provide results for each, so the abstract alone cannot show which designs constrained performance or how broadly the findings apply.","topicIds":["intelligence-and-discovery"],"domain":"science","claimType":"attributed reading","limits":"The evidence provided is the abstract, which does not name the five strategies or give their individual effects. The question set includes faculty-designed questions and questions from online courses; the abstract does not describe how representative they are of assessments beyond this collection. The abstract reports approximate scores for the best-performing model but does not provide exact results for each model, answer format or strategy.","speaker":{"name":"Mingole, Bonam, Yadav, Davis, Lee, Hui Min, Sundar, S. Shyam, Yadav, Amulya","roleAtSource":"Article author"},"source":{"title":"Generative AI and the Crisis of Higher Education Assessments: Examining 'AI-Proofing' Strategies for College-Level Assessments","publisher":"arXiv cs.AI","format":"Article","url":"https://arxiv.org/abs/2611.21476","publishedAt":"2026-11-16","dateBasis":"Date basis not provided."},"evidence":[{"text":"The authors collected 75 questions from authentic college coursework spanning 17 disciplines, contributed by 37 faculty members from ten U.S. universities and one U.K. university.","excerpt":"collecting 75 questions from authentic college coursework assessments, spanning 17 disciplines, which were contributed by 37 faculty members from ten U.S. universities and one U.K. university","url":"https://arxiv.org/abs/2611.21476","relationship":"Source-stated"},{"text":"Of the questions, 61 were faculty-designed as 'LLM-proof' and 14 came from online courses.","excerpt":"Of these, 61 questions were intentionally designed by faculty members to be 'LLM-proof', while the remaining 14 questions were sourced from online courses.","url":"https://arxiv.org/abs/2611.21476","relationship":"Source-stated"},{"text":"The authors evaluated three LLMs with and without supplementary course materials, identified five faculty strategies through thematic analysis, and tested whether those strategies affected answer accuracy.","excerpt":"evaluate three state-of-the-art LLMs' ability to correctly answer these questions with and without supplementary course materials; (ii) identify five strategies, through thematic analysis","url":"https://arxiv.org/abs/2611.21476","relationship":"Source-stated"},{"text":"The best-performing model scored approximately 83% without supplementary course materials and 85% with them.","excerpt":"the best-performing model scored approximately 83% without supplementary course materials and 85% with them","url":"https://arxiv.org/abs/2611.21476","relationship":"Source-stated"},{"text":"The authors observed differences across models, answer formats and instructor strategies.","excerpt":"we observed noteworthy differences across models, answer formats, and instructor tweaking strategies","url":"https://arxiv.org/abs/2611.21476","relationship":"Source-stated"},{"text":"The abstract does not identify the five strategies or report their individual effects, so it does not establish from this evidence which strategies constrained LLM performance.","excerpt":"test whether these strategies impact LLMs' ability to answer the questions accurately","url":"https://arxiv.org/abs/2611.21476","relationship":"Escape Meridian interpretation"}],"sourceMedia":{"status":"not_found","sourceUrl":"https://arxiv.org/abs/2611.21476","checkedAt":"2026-11-17T09:36:08.760Z","items":[],"note":"No suitable media was found by the supported page extraction."},"assumptions":[],"openQuestion":{"attribution":"Open question","text":"Which of the five instructor strategies, if any, meaningfully constrained LLM performance, and how did their effects vary across models and answer formats?"},"whatWouldChangeView":{"attribution":"Change condition not recorded","text":""},"editorialComparisons":[],"live":{"title":"Generative AI and the Crisis of Higher Education Assessments: Examining 'AI-Proofing' Strategies for College-Level Assessments","summary":"In an arXiv abstract, authors Bonam Mingole, Davis Yadav, Hui Min Lee, S. Shyam Sundar and Amulya Yadav describe an academic-integrity challenge: students can outsource homework, take-home assignments and online examinations to large language models (LLMs) that complete them in seconds. They examine whether assessment questions designed to resist LLMs constrain model performance. The authors collected questions from authentic college coursework across disciplines, including faculty-designed questions and questions from online courses, and tested three LLMs with and without supplementary course materials. They report that the best-performing model scored approximately 83% without supplementary materials and 85% with them. They also report differences across models, answer formats and instructor strategies, and say they identified five strategies through thematic analysis. The findings point to a practical tension for instructors: questions intended to make coursework harder for LLMs may not do so equally, and performance varied with the question and model. The abstract says the authors discuss implications for assessment design, academic-integrity policy and the ethics of deploying LLMs in education. It does not name the five strategies or provide results for each, so the abstract alone cannot show which designs constrained performance or how broadly the findings apply.","question":"Which of the five instructor strategies, if any, meaningfully constrained LLM performance, and how did their effects vary across models and answer formats?","claims":[{"text":"The authors collected 75 questions from authentic college coursework spanning 17 disciplines, contributed by 37 faculty members from ten U.S. universities and one U.K. university.","excerpt":"collecting 75 questions from authentic college coursework assessments, spanning 17 disciplines, which were contributed by 37 faculty members from ten U.S. universities and one U.K. university","support":"source-stated"},{"text":"Of the questions, 61 were faculty-designed as 'LLM-proof' and 14 came from online courses.","excerpt":"Of these, 61 questions were intentionally designed by faculty members to be 'LLM-proof', while the remaining 14 questions were sourced from online courses.","support":"source-stated"},{"text":"The authors evaluated three LLMs with and without supplementary course materials, identified five faculty strategies through thematic analysis, and tested whether those strategies affected answer accuracy.","excerpt":"evaluate three state-of-the-art LLMs' ability to correctly answer these questions with and without supplementary course materials; (ii) identify five strategies, through thematic analysis","support":"source-stated"},{"text":"The best-performing model scored approximately 83% without supplementary course materials and 85% with them.","excerpt":"the best-performing model scored approximately 83% without supplementary course materials and 85% with them","support":"source-stated"},{"text":"The authors observed differences across models, answer formats and instructor strategies.","excerpt":"we observed noteworthy differences across models, answer formats, and instructor tweaking strategies","support":"source-stated"},{"text":"The abstract does not identify the five strategies or report their individual effects, so it does not establish from this evidence which strategies constrained LLM performance.","excerpt":"test whether these strategies impact LLMs' ability to answer the questions accurately","support":"interpretation"}],"limitations":["The evidence provided is the abstract, which does not name the five strategies or give their individual effects.","The question set includes faculty-designed questions and questions from online courses; the abstract does not describe how representative they are of assessments beyond this collection.","The abstract reports approximate scores for the best-performing model but does not provide exact results for each model, answer format or strategy."],"mediaId":"ab00ed09a09f35c1b657d28b737e86f3cd07ef2215d3aa481ba10b2a9894b2cf","correctionReason":null,"id":"bb54c984f590d5005339b291baa28f7bdb8a7433094f678651d7b1473ae2227f","publicationId":"9f44839f-ab6f-402e-8dae-91c1f540b0ad","version":1,"sourceUrl":"https://arxiv.org/abs/2611.21476","sourceTitle":"Generative AI and the Crisis of Higher Education Assessments: Examining 'AI-Proofing' Strategies for College-Level Assessments","publisher":"arXiv cs.AI","publishedAt":null,"originalPublished":{"value":"2026-11-16","precision":"day","basis":"publisher_citation_date"},"firstObservedAt":"2026-11-17T08:28:16.131Z","createdAt":"2026-11-17T08:32:02.943Z","historical":false,"entityIds":["arxiv"],"topicIds":["intelligence-and-discovery"],"media":[{"id":"ab00ed09a09f35c1b657d28b737e86f3cd07ef2215d3aa481ba10b2a9894b2cf","kind":"article","url":"https://arxiv.org/abs/2611.21476","sourceUrl":"https://arxiv.org/abs/2611.21476","label":"Read original","provenance":"source_link","verification":{"status":"verified","checkedAt":"2026-11-17T08:30:43.190Z","method":"acquired_source","contentType":"text/html"},"recommendationReason":"Recommended by Jev for understanding this source; checked by Luna."}],"sourceMedia":{"status":"not_found","sourceUrl":"https://arxiv.org/abs/2611.21476","checkedAt":"2026-11-17T09:36:08.760Z","items":[],"note":"No suitable media was found by the supported page extraction."},"editorialDecision":{"outcome":"accept","reasonCodes":["material_development"],"rubricVersion":"stream-admission-v9","mediaId":"ab00ed09a09f35c1b657d28b737e86f3cd07ef2215d3aa481ba10b2a9894b2cf","confidence":0.65,"provider":"typesafe","model":"jev-1.13.0","inputHash":"69f418c043028d0664befb7c462bc1d321ddbc08168124c33e2aca51d63fcb4f","decidedAt":"2026-11-17T08:30:43.879Z"},"provenance":{"textHash":"1832a06040f9dc866ca08f7e6878ce8277871ca6bed1ab9a16fbfeb66bf08e0d","observationId":"2f70114a057fa4eb06dc23147655a8b1ca17a7c58b67ae60a0f3af5184eb4765","retrievedAt":"2026-11-17T08:30:43.082Z","kind":"publisher-page","preparer":{"model":"gpt-6-luna","runId":"b84dc540-8dc9-4887-8140-b3b86edeb341"},"reviewer":{"model":"gpt-6-luna","runId":"0a548ed2-e789-4a30-8792-5819058d00cb"}},"audit":{"status":"passed","reviewedAt":"2026-11-17T08:32:02.943Z","model":"anthropic/claude-opus-5.5"},"sourceAttribution":{"authors":[{"name":"Mingole, Bonam","url":null,"kind":"unknown"},{"name":"Yadav, Davis","url":null,"kind":"unknown"},{"name":"Lee, Hui Min","url":null,"kind":"unknown"},{"name":"Sundar, S. Shyam","url":null,"kind":"unknown"},{"name":"Yadav, Amulya","url":null,"kind":"unknown"}],"authorRole":"article-author","hosts":[],"status":"stated","authorRoster":{"truncated":false,"omitted":0,"assertionGroupsOmitted":0},"participation":[{"actor":{"name":"Mingole, Bonam","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Mingole, Bonam","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2611.21476","locator":"head/meta[name=author|citation_author]"},"assertionId":"39fccf78cf33935498d51564d9a7328774d0812e09a36576025333416f668d18:participation:0","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-17T09:36:08.760Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"39fccf78cf33935498d51564d9a7328774d0812e09a36576025333416f668d18","sourceUrl":"https://arxiv.org/abs/2611.21476","contentSha256":"ae53fafe9106dae3059452ed64d617823f44e9d7997053c0e0c9c019ea3e07ef","textHash":"1832a06040f9dc866ca08f7e6878ce8277871ca6bed1ab9a16fbfeb66bf08e0d","observedAt":"2026-11-17T09:36:08.760Z"}},{"actor":{"name":"Yadav, Davis","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Yadav, Davis","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2611.21476","locator":"head/meta[name=author|citation_author]"},"assertionId":"39fccf78cf33935498d51564d9a7328774d0812e09a36576025333416f668d18:participation:1","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-17T09:36:08.760Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"39fccf78cf33935498d51564d9a7328774d0812e09a36576025333416f668d18","sourceUrl":"https://arxiv.org/abs/2611.21476","contentSha256":"ae53fafe9106dae3059452ed64d617823f44e9d7997053c0e0c9c019ea3e07ef","textHash":"1832a06040f9dc866ca08f7e6878ce8277871ca6bed1ab9a16fbfeb66bf08e0d","observedAt":"2026-11-17T09:36:08.760Z"}},{"actor":{"name":"Lee, Hui Min","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Lee, Hui Min","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2611.21476","locator":"head/meta[name=author|citation_author]"},"assertionId":"39fccf78cf33935498d51564d9a7328774d0812e09a36576025333416f668d18:participation:2","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-17T09:36:08.760Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"39fccf78cf33935498d51564d9a7328774d0812e09a36576025333416f668d18","sourceUrl":"https://arxiv.org/abs/2611.21476","contentSha256":"ae53fafe9106dae3059452ed64d617823f44e9d7997053c0e0c9c019ea3e07ef","textHash":"1832a06040f9dc866ca08f7e6878ce8277871ca6bed1ab9a16fbfeb66bf08e0d","observedAt":"2026-11-17T09:36:08.760Z"}},{"actor":{"name":"Sundar, S. Shyam","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Sundar, S. Shyam","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2611.21476","locator":"head/meta[name=author|citation_author]"},"assertionId":"39fccf78cf33935498d51564d9a7328774d0812e09a36576025333416f668d18:participation:3","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-17T09:36:08.760Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"39fccf78cf33935498d51564d9a7328774d0812e09a36576025333416f668d18","sourceUrl":"https://arxiv.org/abs/2611.21476","contentSha256":"ae53fafe9106dae3059452ed64d617823f44e9d7997053c0e0c9c019ea3e07ef","textHash":"1832a06040f9dc866ca08f7e6878ce8277871ca6bed1ab9a16fbfeb66bf08e0d","observedAt":"2026-11-17T09:36:08.760Z"}},{"actor":{"name":"Yadav, Amulya","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Yadav, Amulya","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2611.21476","locator":"head/meta[name=author|citation_author]"},"assertionId":"39fccf78cf33935498d51564d9a7328774d0812e09a36576025333416f668d18:participation:4","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-17T09:36:08.760Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"39fccf78cf33935498d51564d9a7328774d0812e09a36576025333416f668d18","sourceUrl":"https://arxiv.org/abs/2611.21476","contentSha256":"ae53fafe9106dae3059452ed64d617823f44e9d7997053c0e0c9c019ea3e07ef","textHash":"1832a06040f9dc866ca08f7e6878ce8277871ca6bed1ab9a16fbfeb66bf08e0d","observedAt":"2026-11-17T09:36:08.760Z"}}],"participationCoverage":{"total":5,"returned":5,"complete":true,"fullAssertions":"source-metadata-endpoint"},"distribution":null,"publisher":null,"sourceUrl":"https://arxiv.org/abs/2611.21476","snapshotId":"39fccf78cf33935498d51564d9a7328774d0812e09a36576025333416f668d18","contentSha256":"ae53fafe9106dae3059452ed64d617823f44e9d7997053c0e0c9c019ea3e07ef","textHash":"1832a06040f9dc866ca08f7e6878ce8277871ca6bed1ab9a16fbfeb66bf08e0d","observedAt":"2026-11-17T09:36:08.760Z","basis":"publisher-page-metadata","sourceLinks":43,"linksTruncated":false,"metadataUrl":"/api/pipeline/stream/bb54c984f590d5005339b291baa28f7bdb8a7433094f678651d7b1473ae2227f/source"}}}],"audits":[{"publicationId":"9f44839f-ab6f-402e-8dae-91c1f540b0ad","verdict":"passed","findings":[{"dimension":"attribution","status":"pass","explanation":"The reading names the five listed authors and correctly frames the claims as statements from their arXiv abstract. It does not present the abstract's account of an integrity challenge or its results as independently verified fact.","excerpt":"Authors:Bonam Mingole, Davis Yadav, Hui Min Lee, S. Shyam Sundar, Amulya Yadav"},{"dimension":"summary","status":"pass","explanation":"The summary gives the actor, the problem and the method: three LLMs tested on faculty-written questions, with and without course materials. It reports the headline approximately 83%/85% result and states the main gap, that the strategies and their individual effects are not named. LLM is spelled out. The title and question follow from the abstract. The point that a high score was achieved on a set made mostly of 'LLM-proof' questions appears through the claims rather than directly in the prose. That is a stylistic weakness, not a fidelity failure.","excerpt":"Far less is known about how LLMs perform on college-level assessment questions spanning multiple disciplines, that instructors have purposely designed to be AI-resistant"},{"dimension":"date","status":"pass","explanation":"The original publication date of 2026-11-16 matches the v1 arXiv submission date. The feed date is 2026-11-17, which is later.","excerpt":"[Submitted on 16 Nov 2026]"},{"dimension":"support","status":"pass","explanation":"The question counts, disciplines, faculty, universities, the 61/14 split, the three models, the five strategies and the approximate scores all match the abstract. The 'practical tension' sentence is a hedged inference ('may not do so equally') grounded in the reported differences across strategies and models. The limitations accurately note what the abstract omits. No benefits, costs or forecasts are invented.","excerpt":"we observed noteworthy differences across models, answer formats, and instructor tweaking strategies"},{"dimension":"media","status":"pass","explanation":"The selected media is the verified direct link to the arXiv abstract page, which is the original source. No better source media was found.","excerpt":null}],"createdAt":"2026-11-17T08:32:02.943Z","model":"anthropic/claude-opus-5.5"}],"correctionUrl":"https://escapemeridian.com/corrections/"}