{"id":"https://openalex.org/W7225643282","doi":"https://doi.org/10.21203/rs.3.rs-11641294/v1","title":"Self-Healing JSON Generation in Small Language Models: A Systematic Evaluation of Syntax Repair, Schema Coercion, and Validator Feedback","display_name":"Self-Healing JSON Generation in Small Language Models: A Systematic Evaluation of Syntax Repair, Schema Coercion, and Validator Feedback","publication_year":2026,"publication_date":"2026-11-09","ids":{"openalex":"https://openalex.org/W7225643282","doi":"https://doi.org/10.21203/rs.3.rs-11641294/v1"},"language":"en","primary_location":{"id":"doi:10.21203/rs.3.rs-11641294/v1","doi":"https://doi.org/10.21203/rs.3.rs-11641294/v1","is_oa":true,"landing_page_url":"https://doi.org/10.21203/rs.3.rs-11641294/v1","pdf_url":null,"source":{"id":"https://openalex.org/S4306525896","display_name":"Research Square","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"host_organization_lineage_names":[],"type":"repository"},"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":"Research Square","raw_type":"posted-content"},"volume":null,"type":"preprint","is_traditional":true,"indexed_in":["crossref"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://doi.org/10.21203/rs.3.rs-11641294/v1","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5089030309","display_name":"Ahmed Abdelrahman","orcid":"https://orcid.org/0000-0003-1024-5469","observed_orcids":["https://orcid.org/0000-0003-1024-5469","https://orcid.org/0009-0000-5251-5274","https://orcid.org/0000-0001-5721-8463"]},"institutions":[],"countries":[],"is_corresponding":true,"raw_author_name":"Ahmad Abdelrahman","raw_affiliation_strings":["independet researcher"],"raw_orcid":"https://orcid.org/0009-0000-5251-5274","affiliations":[{"raw_affiliation_string":"independet researcher","institution_ids":[]}]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":0,"corresponding_author_ids":["https://openalex.org/A5089030309"],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":null,"has_fulltext":false,"cited_by_count":0,"citation_normalized_percentile":null,"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":null,"last_page":null},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10181","display_name":"Natural Language Processing Techniques","score":0.5785999894142151,"subfield":{"id":"https://openalex.org/subfields/1702","display_name":"Artificial Intelligence"},"field":{"id":"https://openalex.org/fields/17","display_name":"Computer Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},"topics":[{"id":"https://openalex.org/T10181","display_name":"Natural Language Processing Techniques","score":0.5785999894142151,"subfield":{"id":"https://openalex.org/subfields/1702","display_name":"Artificial Intelligence"},"field":{"id":"https://openalex.org/fields/17","display_name":"Computer Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},{"id":"https://openalex.org/T10028","display_name":"Topic Modeling","score":0.35839998722076416,"subfield":{"id":"https://openalex.org/subfields/1702","display_name":"Artificial Intelligence"},"field":{"id":"https://openalex.org/fields/17","display_name":"Computer Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},{"id":"https://openalex.org/T12031","display_name":"Speech and dialogue systems","score":0.013799999840557575,"subfield":{"id":"https://openalex.org/subfields/1702","display_name":"Artificial Intelligence"},"field":{"id":"https://openalex.org/fields/17","display_name":"Computer Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}}],"keywords":[],"concepts":[{"id":"https://openalex.org/C35292069","wikidata":"https://www.wikidata.org/wiki/Q1575458","display_name":"Validator","level":2,"score":0.9642000198364258},{"id":"https://openalex.org/C2780416260","wikidata":"https://www.wikidata.org/wiki/Q2063","display_name":"JSON","level":2,"score":0.9467999935150146},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.8414000272750854},{"id":"https://openalex.org/C186644900","wikidata":"https://www.wikidata.org/wiki/Q194152","display_name":"Parsing","level":2,"score":0.6809999942779541},{"id":"https://openalex.org/C52146309","wikidata":"https://www.wikidata.org/wiki/Q7431116","display_name":"Schema (genetic algorithms)","level":2,"score":0.579200029373169},{"id":"https://openalex.org/C199360897","wikidata":"https://www.wikidata.org/wiki/Q9143","display_name":"Programming language","level":1,"score":0.5641000270843506},{"id":"https://openalex.org/C88006597","wikidata":"https://www.wikidata.org/wiki/Q690117","display_name":"Disk formatting","level":2,"score":0.39430001378059387},{"id":"https://openalex.org/C204321447","wikidata":"https://www.wikidata.org/wiki/Q30642","display_name":"Natural language processing","level":1,"score":0.3882000148296356},{"id":"https://openalex.org/C60048249","wikidata":"https://www.wikidata.org/wiki/Q37437","display_name":"Syntax","level":2,"score":0.3846000134944916},{"id":"https://openalex.org/C544833334","wikidata":"https://www.wikidata.org/wiki/Q2005","display_name":"JavaScript","level":2,"score":0.38019999861717224},{"id":"https://openalex.org/C43126263","wikidata":"https://www.wikidata.org/wiki/Q128751","display_name":"Source code","level":2,"score":0.3596000075340271},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.3490000069141388},{"id":"https://openalex.org/C150921843","wikidata":"https://www.wikidata.org/wiki/Q1170431","display_name":"Resampling","level":2,"score":0.34299999475479126},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.33500000834465027},{"id":"https://openalex.org/C43521106","wikidata":"https://www.wikidata.org/wiki/Q2165493","display_name":"Pipeline (software)","level":2,"score":0.32749998569488525},{"id":"https://openalex.org/C68699486","wikidata":"https://www.wikidata.org/wiki/Q265904","display_name":"Document Structure Description","level":3,"score":0.3061999976634979},{"id":"https://openalex.org/C26022165","wikidata":"https://www.wikidata.org/wiki/Q8091","display_name":"Grammar","level":2,"score":0.3018999993801117},{"id":"https://openalex.org/C133162039","wikidata":"https://www.wikidata.org/wiki/Q1061077","display_name":"Code generation","level":3,"score":0.2948000133037567},{"id":"https://openalex.org/C176982825","wikidata":"https://www.wikidata.org/wiki/Q835922","display_name":"Lexical analysis","level":2,"score":0.2842000126838684},{"id":"https://openalex.org/C58646249","wikidata":"https://www.wikidata.org/wiki/Q127380","display_name":"Abstract syntax tree","level":3,"score":0.26910001039505005},{"id":"https://openalex.org/C23123220","wikidata":"https://www.wikidata.org/wiki/Q816826","display_name":"Information retrieval","level":1,"score":0.26440000534057617},{"id":"https://openalex.org/C2776760102","wikidata":"https://www.wikidata.org/wiki/Q5139990","display_name":"Code (set theory)","level":3,"score":0.2635999917984009},{"id":"https://openalex.org/C12590798","wikidata":"https://www.wikidata.org/wiki/Q3933199","display_name":"Replication (statistics)","level":2,"score":0.2558000087738037}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.21203/rs.3.rs-11641294/v1","doi":"https://doi.org/10.21203/rs.3.rs-11641294/v1","is_oa":true,"landing_page_url":"https://doi.org/10.21203/rs.3.rs-11641294/v1","pdf_url":null,"source":{"id":"https://openalex.org/S4306525896","display_name":"Research Square","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"host_organization_lineage_names":[],"type":"repository"},"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":"Research Square","raw_type":"posted-content"}],"best_oa_location":{"id":"doi:10.21203/rs.3.rs-11641294/v1","doi":"https://doi.org/10.21203/rs.3.rs-11641294/v1","is_oa":true,"landing_page_url":"https://doi.org/10.21203/rs.3.rs-11641294/v1","pdf_url":null,"source":{"id":"https://openalex.org/S4306525896","display_name":"Research Square","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"host_organization_lineage_names":[],"type":"repository"},"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":"Research Square","raw_type":"posted-content"},"sustainable_development_goals":[],"study_designs":[],"awards":[],"funders":[],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":7,"referenced_works":["https://openalex.org/W4385014449","https://openalex.org/W4388718089","https://openalex.org/W4405023514","https://openalex.org/W4405425153","https://openalex.org/W4405655184","https://openalex.org/W4406733450","https://openalex.org/W4407209014"],"related_works":[],"abstract_inverted_index":{"Abstract":[0],"Small":[1],"language":[2],"models":[3,107],"offer":[4],"an":[5,147],"affordable":[6],"way":[7],"to":[8,88,111,134,173],"extract":[9],"structured":[10],"information,":[11],"including":[12],"on":[13,43,95,207],"devices":[14],"with":[15,146,224],"limited":[16],"computing":[17],"resources.":[18],"However,":[19],"their":[20],"responses":[21,124,267],"often":[22,188],"contain":[23,269],"JSON":[24,84,164,257],"that":[25,41,72,82],"cannot":[26],"be":[27],"parsed":[28],"or":[29,49,193,275],"does":[30,217,258],"not":[31,218,247,259],"satisfy":[32,125],"the":[33,51,74,89,126,177,190,195,236,243],"required":[34],"schema.":[35,127],"We":[36,91,282],"investigate":[37],"a":[38,60,79,228,278],"self-healing":[39],"approach":[40,55],"operates":[42],"generated":[44],"text":[45],"alone,":[46],"without":[47,136,175],"changing":[48],"accessing":[50],"model's":[52],"decoder.":[53],"The":[54,140,154],"combines":[56],"three":[57,105],"steps:":[58],"(L1)":[59],"tolerant":[61],"lexer":[62],"and":[63,77,114,130,289],"bracket-stack":[64],"parser":[65],"for":[66,235],"syntax":[67],"repair,":[68],"(L2)":[69],"schema-guided":[70],"conversions":[71],"preserve":[73],"original":[75],"meaning,":[76],"(L3)":[78],"retry":[80],"mechanism":[81],"provides":[83],"Schema":[85],"validation":[86],"errors":[87],"model.":[90],"test":[92],"these":[93,204,294],"steps":[94],"120":[96],"seeded":[97],"text-to-JSON":[98],"tasks":[99],"spanning":[100],"four":[101],"schema":[102,196],"families,":[103],"using":[104],"instruction-tuned":[106],"ranging":[108],"from":[109],"360M":[110],"1.5B":[112],"parameters":[113],"six":[115],"experimental":[116],"conditions.":[117],"On":[118],"average,":[119],"only":[120],"16.7%":[121],"of":[122,149,264],"raw":[123],"Applying":[128],"L1":[129],"L2":[131],"raises":[132],"validity":[133,145],"67.8%":[135],"another":[137],"model":[138,178,238],"call.":[139],"complete":[141],"pipeline":[142],"achieves":[143],"78.1%":[144],"average":[148],"1.57":[150],"calls":[151],"per":[152],"task.":[153],"results":[155,234],"reveal":[156],"several":[157],"important":[158],"distinctions.":[159],"Qwen2.5":[160],"frequently":[161],"produces":[162],"correct":[163,174],"inside":[165],"Markdown":[166],"code":[167,185],"fences,":[168],"making":[169],"many":[170],"failures":[171],"easy":[172],"involving":[176],"again.":[179],"SmolLM2-360M,":[180],"by":[181],"contrast,":[182],"generally":[183],"avoids":[184],"fences":[186],"but":[187],"misunderstands":[189],"requested":[191],"structure":[192],"copies":[194],"itself;":[197],"validator":[198],"feedback":[199,232],"is":[200,246],"more":[201,279],"useful":[202],"in":[203],"cases.":[205],"Feedback":[206],"its":[208],"own":[209],"performs":[210],"poorly":[211],"(21.1%":[212],"validity),":[213],"largely":[214],"because":[215],"it":[216],"break":[219],"persistent":[220],"formatting":[221],"habits.":[222],"Compared":[223],"blind":[225],"resampling":[226],"at":[227,270],"similar":[229],"call":[230],"budget,":[231],"improves":[233],"weakest":[237],"(p":[239,250],"=":[240,251],"0.003),":[241],"although":[242],"overall":[244],"difference":[245],"statistically":[248],"significant":[249],"0.060).":[252],"More":[253],"importantly,":[254],"producing":[255],"valid":[256,265],"guarantee":[260],"accurate":[261],"information:":[262],"45.6%":[263],"self-healed":[266],"still":[268],"least":[271],"one":[272],"incorrect":[273],"field,":[274],"38.4%":[276],"under":[277],"forgiving":[280],"comparison.":[281],"observe":[283],"empty":[284],"strings,":[285],"fabricated":[286],"phone":[287],"numbers,":[288],"default":[290],"enum":[291],"choices":[292],"among":[293],"errors.":[295]},"counts_by_year":[],"updated_date":"2026-11-10T13:44:29.811591","created_date":"2026-11-10T00:00:00"}