{"id":"https://openalex.org/W7163369080","doi":"https://doi.org/10.48550/arxiv.2606.03130","title":"Synthetic Hallucinations, Real Gains: Hard Negatives from Frontier Models for FIM Hallucination Mitigation","display_name":"Synthetic Hallucinations, Real Gains: Hard Negatives from Frontier Models for FIM Hallucination Mitigation","publication_year":2026,"publication_date":"2026-06-02","ids":{"openalex":"https://openalex.org/W7163369080","doi":"https://doi.org/10.48550/arxiv.2606.03130"},"language":"en","primary_location":{"id":"pmh:oai:arXiv.org:2606.03130","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2606.03130","pdf_url":"https://arxiv.org/pdf/2606.03130","source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":"https://openalex.org/I205783295","host_organization_name":"Cornell University","host_organization_lineage":["https://openalex.org/I205783295"],"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":null,"raw_type":"text"},"volume":null,"type":"preprint","is_traditional":true,"indexed_in":["arxiv","datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://arxiv.org/pdf/2606.03130","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5111162538","display_name":"Mahdi Erfanian","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I39422238","display_name":"University of Illinois Chicago","ror":"https://ror.org/02mpq6x41","country_code":"US","type":"education","lineage":["https://openalex.org/I39422238"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Erfanian, Mahdi","raw_affiliation_strings":["University of Illinois Chicago"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"University of Illinois Chicago","institution_ids":["https://openalex.org/I39422238"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5133675173","display_name":"Nelson Daniel Troncoso","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Troncoso, Nelson Daniel","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5044348902","display_name":"Aashna Garg","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Garg, Aashna","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5135867672","display_name":"Amabel Gale","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I1290206253","display_name":"Microsoft (United States)","ror":"https://ror.org/00d0nc645","country_code":"US","type":"company","lineage":["https://openalex.org/I1290206253"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Gale, Amabel","raw_affiliation_strings":["Microsoft"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Microsoft","institution_ids":["https://openalex.org/I1290206253"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5153511247","display_name":"Xiaoyu Liu","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I1290206253","display_name":"Microsoft (United States)","ror":"https://ror.org/00d0nc645","country_code":"US","type":"company","lineage":["https://openalex.org/I1290206253"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Liu, Xiaoyu","raw_affiliation_strings":["Microsoft"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Microsoft","institution_ids":["https://openalex.org/I1290206253"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5041554597","display_name":"Pareesa Ameneh Golnari","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I1290206253","display_name":"Microsoft (United States)","ror":"https://ror.org/00d0nc645","country_code":"US","type":"company","lineage":["https://openalex.org/I1290206253"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Golnari, Pareesa Ameneh","raw_affiliation_strings":["Microsoft"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Microsoft","institution_ids":["https://openalex.org/I1290206253"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5135827465","display_name":"Shengyu Fu","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Fu, Shengyu","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":2,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":null,"has_fulltext":true,"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/T10260","display_name":"Software Engineering Research","score":0.9980000257492065,"subfield":{"id":"https://openalex.org/subfields/1710","display_name":"Information Systems"},"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/T10260","display_name":"Software Engineering Research","score":0.9980000257492065,"subfield":{"id":"https://openalex.org/subfields/1710","display_name":"Information Systems"},"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":0.0006000000284984708,"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/T10181","display_name":"Natural Language Processing Techniques","score":0.00019999999494757503,"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":[{"id":"https://openalex.org/keywords/hallucination-mitigation","display_name":"hallucination mitigation","score":0.7760000228881836},{"id":"https://openalex.org/keywords/hard-negatives","display_name":"hard negatives","score":0.6370000243186951},{"id":"https://openalex.org/keywords/code-generation","display_name":"code generation","score":0.5400000214576721},{"id":"https://openalex.org/keywords/fine-tuning","display_name":"fine-tuning","score":0.6159999966621399},{"id":"https://openalex.org/keywords/qwen2-5-coder","display_name":"Qwen2.5-Coder","score":0.8140000104904175}],"concepts":[{"id":"https://openalex.org/C2911011789","wikidata":"https://www.wikidata.org/wiki/Q130741","display_name":"Hallucinating","level":2,"score":0.6942999958992004},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6840999722480774},{"id":"https://openalex.org/C43521106","wikidata":"https://www.wikidata.org/wiki/Q2165493","display_name":"Pipeline (software)","level":2,"score":0.6244999766349792},{"id":"https://openalex.org/C2779343474","wikidata":"https://www.wikidata.org/wiki/Q3109175","display_name":"Context (archaeology)","level":2,"score":0.5877000093460083},{"id":"https://openalex.org/C153083717","wikidata":"https://www.wikidata.org/wiki/Q6535263","display_name":"Leverage (statistics)","level":2,"score":0.5529000163078308},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.5209000110626221},{"id":"https://openalex.org/C2776760102","wikidata":"https://www.wikidata.org/wiki/Q5139990","display_name":"Code (set theory)","level":3,"score":0.5159000158309937},{"id":"https://openalex.org/C204321447","wikidata":"https://www.wikidata.org/wiki/Q30642","display_name":"Natural language processing","level":1,"score":0.48330000042915344},{"id":"https://openalex.org/C28719098","wikidata":"https://www.wikidata.org/wiki/Q44946","display_name":"Point (geometry)","level":2,"score":0.41499999165534973},{"id":"https://openalex.org/C103278499","wikidata":"https://www.wikidata.org/wiki/Q254465","display_name":"Similarity (geometry)","level":3,"score":0.3571000099182129},{"id":"https://openalex.org/C51929080","wikidata":"https://www.wikidata.org/wiki/Q2425187","display_name":"Codebase","level":3,"score":0.3450999855995178},{"id":"https://openalex.org/C43126263","wikidata":"https://www.wikidata.org/wiki/Q128751","display_name":"Source code","level":2,"score":0.33889999985694885},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.33820000290870667},{"id":"https://openalex.org/C195324797","wikidata":"https://www.wikidata.org/wiki/Q33742","display_name":"Natural language","level":2,"score":0.33149999380111694},{"id":"https://openalex.org/C2994044699","wikidata":"https://www.wikidata.org/wiki/Q1330121","display_name":"First pass","level":2,"score":0.31310001015663147},{"id":"https://openalex.org/C150140777","wikidata":"https://www.wikidata.org/wiki/Q960648","display_name":"Point of interest","level":2,"score":0.2976999878883362},{"id":"https://openalex.org/C2776650193","wikidata":"https://www.wikidata.org/wiki/Q264661","display_name":"Obstacle","level":2,"score":0.29089999198913574},{"id":"https://openalex.org/C199360897","wikidata":"https://www.wikidata.org/wiki/Q9143","display_name":"Programming language","level":1,"score":0.28600001335144043},{"id":"https://openalex.org/C160920958","wikidata":"https://www.wikidata.org/wiki/Q7662746","display_name":"Synthetic data","level":2,"score":0.26910001039505005},{"id":"https://openalex.org/C48145219","wikidata":"https://www.wikidata.org/wiki/Q1335365","display_name":"Security token","level":2,"score":0.2676999866962433},{"id":"https://openalex.org/C2164484","wikidata":"https://www.wikidata.org/wiki/Q5170150","display_name":"Core (optical fiber)","level":2,"score":0.2660999894142151},{"id":"https://openalex.org/C124304363","wikidata":"https://www.wikidata.org/wiki/Q673661","display_name":"Abstraction","level":2,"score":0.26510000228881836},{"id":"https://openalex.org/C2778571376","wikidata":"https://www.wikidata.org/wiki/Q1355821","display_name":"Frontier","level":2,"score":0.2635999917984009},{"id":"https://openalex.org/C175309249","wikidata":"https://www.wikidata.org/wiki/Q725864","display_name":"Pipeline transport","level":2,"score":0.2596000134944916},{"id":"https://openalex.org/C2776502983","wikidata":"https://www.wikidata.org/wiki/Q690182","display_name":"Contrast (vision)","level":2,"score":0.25459998846054077},{"id":"https://openalex.org/C31972630","wikidata":"https://www.wikidata.org/wiki/Q844240","display_name":"Computer vision","level":1,"score":0.25440001487731934}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:oai:arXiv.org:2606.03130","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2606.03130","pdf_url":"https://arxiv.org/pdf/2606.03130","source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":"https://openalex.org/I205783295","host_organization_name":"Cornell University","host_organization_lineage":["https://openalex.org/I205783295"],"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":null,"raw_type":"text"},{"id":"doi:10.48550/arxiv.2606.03130","doi":"https://doi.org/10.48550/arxiv.2606.03130","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2606.03130","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":"https://openalex.org/I205783295","host_organization_name":"Cornell University","host_organization_lineage":["https://openalex.org/I205783295"],"host_organization_lineage_names":[],"type":"repository"},"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":null,"is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"Preprint"}],"best_oa_location":{"id":"pmh:oai:arXiv.org:2606.03130","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2606.03130","pdf_url":"https://arxiv.org/pdf/2606.03130","source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":"https://openalex.org/I205783295","host_organization_name":"Cornell University","host_organization_lineage":["https://openalex.org/I205783295"],"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":null,"raw_type":"text"},"sustainable_development_goals":[],"study_designs":[],"awards":[],"funders":[],"has_content":{"pdf":true,"grobid_xml":true},"content_urls":{"pdf":"https://content.openalex.org/works/W7163369080.pdf","grobid_xml":"https://content.openalex.org/works/W7163369080.grobid-xml"},"referenced_works_count":18,"referenced_works":["https://openalex.org/W4224060952","https://openalex.org/W4281669078","https://openalex.org/W4288804596","https://openalex.org/W4320858785","https://openalex.org/W4385572345","https://openalex.org/W4385572634","https://openalex.org/W4391272793","https://openalex.org/W4392677872","https://openalex.org/W4393969224","https://openalex.org/W4396570275","https://openalex.org/W4402667093","https://openalex.org/W4404782689","https://openalex.org/W4409362569","https://openalex.org/W4414888755","https://openalex.org/W6967084484","https://openalex.org/W7120088840","https://openalex.org/W7133208539","https://openalex.org/W7160872720"],"related_works":["https://openalex.org/W7160872720","https://openalex.org/W4416258724","https://openalex.org/W4407953060","https://openalex.org/W7125204071","https://openalex.org/W4392677872","https://openalex.org/W7166888659","https://openalex.org/W4390601895","https://openalex.org/W4414888755","https://openalex.org/W4401635661","https://openalex.org/W7083818868"],"abstract_inverted_index":{"Small":[0],"open-source":[1],"code":[2,59,220],"models":[3,60],"that":[4,23,38,47,234],"power":[5],"IDE":[6],"autocomplete":[7],"still":[8],"emit":[9],"hallucinated":[10],"Fill-in-the-Middle":[11],"(FIM)":[12],"completions:":[13],"syntactically":[14],"natural":[15],"calls":[16],"to":[17,61,106],"methods,":[18],"parameters,":[19],"variables,":[20],"and":[21,76,98,149,157,166,195,227],"imports":[22],"do":[24,39],"not":[25,40],"exist":[26],"in":[27,237],"the":[28,70,77,121,212,216,228,235],"surrounding":[29],"project.":[30],"Existing":[31],"mitigations":[32],"either":[33],"require":[34],"per-language":[35],"execution":[36],"sandboxes":[37],"apply":[41],"at":[42,173],"mid-keystroke":[43],"or":[44],"preference-optimisation":[45],"pipelines":[46],"need":[48],"large":[49],"human-labelled":[50],"corpora.":[51],"We":[52,214],"propose":[53],"an":[54],"execution-free":[55],"alternative:":[56],"use":[57],"frontier":[58,104],"synthesise":[62],"plausible-but-wrong":[63],"completions":[64],"as":[65,81],"hard":[66,109],"negatives,":[67],"then":[68],"leverage":[69],"contrast":[71],"between":[72],"these":[73],"synthetic":[74],"hallucinations":[75],"ground-truth":[78],"developer":[79],"edit":[80,150],"a":[82,100,124,131,138,181,196,201],"supervised":[83],"fine-tuning":[84,230],"signal.":[85],"Our":[86],"pipeline":[87,218],"scrapes":[88],"multilingual":[89,126],"FIM":[90,127,229],"contexts":[91],"from":[92,120],"public":[93],"GitHub":[94],"across":[95],"eight":[96],"languages":[97],"asks":[99],"panel":[101],"of":[102,115],"three":[103],"generators":[105],"produce":[107],"one":[108],"negative":[110],"per":[111],"context":[112],"for":[113],"each":[114],"four":[116],"hallucination":[117,128],"types":[118],"drawn":[119],"Delulu":[122,143,176],"taxonomy,":[123],"Docker-verified":[125],"benchmark,":[129],"yielding":[130],"paired":[132],"chosen/rejected":[133],"dataset.":[134],"Fine-tuning":[135],"Qwen2.5-Coder-7B-Instruct":[136],"on":[137,154,245],"100K-row":[139],"curated":[140],"subset":[141],"lifts":[142,175],"exact":[144],"match":[145],"by":[146,152,177],"+18.8":[147],"points":[148],"similarity":[151],"+0.22":[153],"every":[155,158,163,167],"language":[156,191],"type,":[159],"while":[160],"also":[161],"improving":[162],"HumanEval-Infilling":[164],"split":[165],"SAFIM":[168],"subset.":[169],"The":[170],"same":[171],"recipe":[172,231],"3B":[174],"+12.8":[178],"EM":[179],"with":[180],"small,":[182],"characterised":[183],"general-FIM":[184],"trade-off.":[185],"Five-axis":[186],"ablations":[187],"(size,":[188],"type":[189],"mix,":[190],"coverage,":[192],"base-model":[193],"family,":[194],"difficulty-aware":[197],"fool":[198],"rate)":[199],"plus":[200],"head-to-head":[202],"SFT":[203],"vs.":[204],"DPO/ORPO":[205],"comparison":[206],"map":[207],"which":[208],"design":[209],"choices":[210],"drive":[211],"gain.":[213],"release":[215],"full":[217],"source":[219],"--":[221,232],"generation,":[222],"fool-rate":[223],"LLM":[224],"judging,":[225],"curation,":[226],"so":[233],"experiments":[236],"this":[238],"paper":[239],"can":[240],"be":[241],"reproduced":[242],"end-to":[243],"end":[244],"any":[246],"permissively":[247],"licensed":[248],"corpus.":[249]},"counts_by_year":[],"updated_date":"2026-11-07T11:25:07.637180","created_date":"2026-06-04T00:00:00"}