Safety and Alignment of Small LanguageModels: A Systematic Survey of Methods,Evaluation, and Open Challenges

Safety and Alignment of Small LanguageModels: A Systematic Survey of Methods,Evaluation, and Open Challenges | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Safety and Alignment of Small LanguageModels: A Systematic Survey of Methods,Evaluation, and Open Challenges Robert-Mihai Colca This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-9557790/v1 This work is licensed under a CC BY 4.0 License Status: Under Review Version 1 posted 9 You are reading this latest preprint version Abstract The rapid proliferation of small language models (SLMs)—defined here as modelswith up to 7 billion active parameters—has democratized access to capable lan-guage technology, allowing deployment on consumer hardware, mobile devices, andedge infrastructure. However, the safety and alignment properties of these modelsremain critically understudied relative to their frontier counterparts. This surveyoffers the first dedicated review of safety and alignment for small language mod-els. We organize the field along seven axes: (1) data-centric safety approaches,(2) training-time alignment techniques and their comparative effectiveness, (3) post-training safeguards and guardrails, (4) model editing and knowledge management,(5) the interaction between efficiency techniques and safety, (6) multilingual safetydisparities, and (7) mechanistic interpretability of safety behaviors. For each axis,we provide a critical comparative analysis identifying strengths, weaknesses, andtrade-offs. We review more than 130 articles published between 2022 and 2026,construct a unified taxonomy, survey evaluation benchmarks and their limitations,catalog known failure modes, and identify eight concrete open research challenges.Our analysis reveals that safety alignment in SLMs is not simply a scaled-downversion of frontier model alignment but presents qualitatively distinct challenges—including acute capacity–safety trade-offs, compression-induced safety degradation,and amplified multilingual vulnerabilities—that require dedicated investigation. Small language models AI safety Alignment Safety evaluation Mechanistic interpretability Multilingual safety Full Text Additional Declarations No competing interests reported. Cite Share Download PDF Status: Under Review Version 1 posted Reviewers agreed at journal 18 May, 2026 Reviews received at journal 14 May, 2026 Reviews received at journal 14 May, 2026 Reviewers agreed at journal 07 May, 2026 Reviewers agreed at journal 05 May, 2026 Reviewers invited by journal 05 May, 2026 Editor assigned by journal 01 May, 2026 Submission checks completed at journal 28 Apr, 2026 First submitted to journal 28 Apr, 2026 You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-9557790","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":636299198,"identity":"c834eb88-8c8b-4c9f-a9b0-0191d0de4036","order_by":0,"name":"Robert-Mihai Colca","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA0UlEQVRIiWNgGAWjYFACHijNDsQfiNeSAMTMDAyMM0ACbKRoYeYhRot5e+/Bx5U/GOT5mZkffrapYEjccL+B7QM+F8qcOZdseCaBwXBmM5uxdM4ZoJZjDMwzZ+DRIiGRYybZkMCQYHCYh0E6t+1/4sw2BmZmHvxazH9CtTD/tvzHQJQWM0aoFjZpxgaGxH42Qlp4ziVLNqRJgPxiZtlzjMG4ny2xmRGvX9h7D35ssLGR52dvfnzjRw2DbBvz4cPExKkEMgfovFEwCkbBKBgFlAEAKLg9cOIevhYAAAAASUVORK5CYII=","orcid":"","institution":"West University of Timisoara","correspondingAuthor":true,"prefix":"","firstName":"Robert-Mihai","middleName":"","lastName":"Colca","suffix":""}],"badges":[],"createdAt":"2026-04-28 19:23:55","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-9557790/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-9557790/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":109053973,"identity":"7b296123-b672-41fa-b7a1-2aebfa03580d","added_by":"auto","created_at":"2026-05-12 07:16:37","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":442270,"visible":true,"origin":"","legend":"","description":"","filename":"submission.pdf","url":"https://assets-eu.researchsquare.com/files/rs-9557790/v1_covered_a338b1e3-2540-4a15-bef4-232c9ccc2610.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"\u003cp\u003eSafety and Alignment of Small LanguageModels: A Systematic Survey of Methods,Evaluation, and Open Challenges\u003c/p\u003e","fulltext":[],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":false,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":true,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":true,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"artificial-intelligence-review","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"aire","sideBox":"Learn more about [Artificial Intelligence Review](http://link.springer.com/journal/10462)","snPcode":"10462","submissionUrl":"https://submission.nature.com/new-submission/10462/3","title":"Artificial Intelligence Review","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"em","reportingPortfolio":"Springer Hybrid","inReviewEnabled":true,"inReviewRevisionsEnabled":false},"keywords":"Small language models, AI safety, Alignment, Safety evaluation, Mechanistic interpretability, Multilingual safety","lastPublishedDoi":"10.21203/rs.3.rs-9557790/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-9557790/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"The rapid proliferation of small language models (SLMs)—defined here as modelswith up to 7 billion active parameters—has democratized access to capable lan-guage technology, allowing deployment on consumer hardware, mobile devices, andedge infrastructure. However, the safety and alignment properties of these modelsremain critically understudied relative to their frontier counterparts. This surveyoffers the first dedicated review of safety and alignment for small language mod-els. We organize the field along seven axes: (1) data-centric safety approaches,(2) training-time alignment techniques and their comparative effectiveness, (3) post-training safeguards and guardrails, (4) model editing and knowledge management,(5) the interaction between efficiency techniques and safety, (6) multilingual safetydisparities, and (7) mechanistic interpretability of safety behaviors. For each axis,we provide a critical comparative analysis identifying strengths, weaknesses, andtrade-offs. We review more than 130 articles published between 2022 and 2026,construct a unified taxonomy, survey evaluation benchmarks and their limitations,catalog known failure modes, and identify eight concrete open research challenges.Our analysis reveals that safety alignment in SLMs is not simply a scaled-downversion of frontier model alignment but presents qualitatively distinct challenges—including acute capacity–safety trade-offs, compression-induced safety degradation,and amplified multilingual vulnerabilities—that require dedicated investigation.","manuscriptTitle":"Safety and Alignment of Small LanguageModels: A Systematic Survey of Methods,Evaluation, and Open Challenges","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-05-12 07:11:55","doi":"10.21203/rs.3.rs-9557790/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"reviewerAgreed","content":"21330176254530791486685237751381215024","date":"2026-05-18T11:26:39+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2026-05-15T02:04:33+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2026-05-14T05:45:03+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"336098822047544822820999708228468481529","date":"2026-05-07T06:31:08+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"332194469148306420654211259217121353235","date":"2026-05-05T06:28:18+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2026-05-05T05:49:37+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2026-05-01T14:11:18+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2026-04-29T01:49:37+00:00","index":"","fulltext":""},{"type":"submitted","content":"Artificial Intelligence Review","date":"2026-04-28T19:10:41+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"artificial-intelligence-review","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"aire","sideBox":"Learn more about [Artificial Intelligence Review](http://link.springer.com/journal/10462)","snPcode":"10462","submissionUrl":"https://submission.nature.com/new-submission/10462/3","title":"Artificial Intelligence Review","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"em","reportingPortfolio":"Springer Hybrid","inReviewEnabled":true,"inReviewRevisionsEnabled":false}}],"origin":"","ownerIdentity":"006bd564-63da-44f5-8019-b52c7d116d87","owner":[],"postedDate":"May 12th, 2026","published":true,"recentEditorialEvents":[{"type":"reviewerAgreed","content":"21330176254530791486685237751381215024","date":"2026-05-18T11:26:39+00:00","index":51,"fulltext":""},{"type":"editorInvitedReview","content":"","date":"2026-05-15T02:04:33+00:00","index":31,"fulltext":""},{"type":"editorInvitedReview","content":"","date":"2026-05-14T05:45:03+00:00","index":30,"fulltext":""},{"type":"reviewerAgreed","content":"336098822047544822820999708228468481529","date":"2026-05-07T06:31:08+00:00","index":25,"fulltext":""},{"type":"reviewerAgreed","content":"332194469148306420654211259217121353235","date":"2026-05-05T06:28:18+00:00","index":24,"fulltext":""},{"type":"reviewersInvited","content":"39","date":"2026-05-05T05:49:37+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2026-05-01T14:11:18+00:00","index":"","fulltext":""}],"rejectedJournal":[],"revision":"","amendment":"","status":"under-review","subjectAreas":[],"tags":[],"updatedAt":"2026-05-12T07:11:55+00:00","versionOfRecord":[],"versionCreatedAt":"2026-05-12 07:11:55","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-9557790","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-9557790","identity":"rs-9557790","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}