{"id":"https://openalex.org/W7214717067","doi":"https://doi.org/10.48550/arxiv.2609.33845","title":"How code helps different tasks? A decompositional lens on LLM post-training","display_name":"How code helps different tasks? A decompositional lens on LLM post-training","publication_year":2026,"publication_date":"2026-09-27","ids":{"openalex":"https://openalex.org/W7214717067","doi":"https://doi.org/10.48550/arxiv.2609.33845"},"language":"en","primary_location":{"id":"pmh:oai:arXiv.org:2609.33845","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2609.33845","pdf_url":"https://arxiv.org/pdf/2609.33845","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":null,"license_id":null,"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/2609.33845","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5153446096","display_name":"Zheng Yu","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I157773358","display_name":"Sun Yat-sen University","ror":"https://ror.org/0064kty71","country_code":"CN","type":"education","lineage":["https://openalex.org/I157773358"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Yu, Zheng","raw_affiliation_strings":["Shenzhen Loop Area Institute","Sun Yat-sen University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Shenzhen Loop Area Institute","institution_ids":[]},{"raw_affiliation_string":"Sun Yat-sen University","institution_ids":["https://openalex.org/I157773358"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5153454508","display_name":"Yiwei Li","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I4210116924","display_name":"Chinese University of Hong Kong, Shenzhen","ror":"https://ror.org/02d5ks197","country_code":"CN","type":"education","lineage":["https://openalex.org/I177725633","https://openalex.org/I180726961","https://openalex.org/I4210116924"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Li, Yiwei","raw_affiliation_strings":["The Chinese University of Hong Kong , Shenzhen"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"The Chinese University of Hong Kong , Shenzhen","institution_ids":["https://openalex.org/I4210116924"]}]},{"author_position":"middle","author":{"id":null,"display_name":"Chen, Yishen","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I4210116924","display_name":"Chinese University of Hong Kong, Shenzhen","ror":"https://ror.org/02d5ks197","country_code":"CN","type":"education","lineage":["https://openalex.org/I177725633","https://openalex.org/I180726961","https://openalex.org/I4210116924"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Chen, Yishen","raw_affiliation_strings":["The Chinese University of Hong Kong , Shenzhen"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"The Chinese University of Hong Kong , Shenzhen","institution_ids":["https://openalex.org/I4210116924"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5153489312","display_name":"Xiang Li","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I4210099586","display_name":"Shenzhen Research Institute of Big Data","ror":"https://ror.org/00z1gwf89","country_code":"CN","type":"education","lineage":["https://openalex.org/I4210099586"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Li, Xiang","raw_affiliation_strings":["Shenzhen Research Institute of Big Data"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Shenzhen Research Institute of Big Data","institution_ids":["https://openalex.org/I4210099586"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5011412438","display_name":"Jiale Han","orcid":"https://orcid.org/0000-0001-6477-0424","observed_orcids":["https://orcid.org/0000-0001-6477-0424"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Han, Jiale","raw_affiliation_strings":["Shenzhen Loop Area Institute"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Shenzhen Loop Area Institute","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5153552082","display_name":"Benyou Wang","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I4210116924","display_name":"Chinese University of Hong Kong, Shenzhen","ror":"https://ror.org/02d5ks197","country_code":"CN","type":"education","lineage":["https://openalex.org/I177725633","https://openalex.org/I180726961","https://openalex.org/I4210116924"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Wang, Benyou","raw_affiliation_strings":["The Chinese University of Hong Kong , Shenzhen"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"The Chinese University of Hong Kong , Shenzhen","institution_ids":["https://openalex.org/I4210116924"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5153416483","display_name":"Jingbang Chen","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I4210116924","display_name":"Chinese University of Hong Kong, Shenzhen","ror":"https://ror.org/02d5ks197","country_code":"CN","type":"education","lineage":["https://openalex.org/I177725633","https://openalex.org/I180726961","https://openalex.org/I4210116924"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Chen, Jingbang","raw_affiliation_strings":["Shenzhen Loop Area Institute","The Chinese University of Hong Kong , Shenzhen"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Shenzhen Loop Area Institute","institution_ids":[]},{"raw_affiliation_string":"The Chinese University of Hong Kong , Shenzhen","institution_ids":["https://openalex.org/I4210116924"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":3,"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/T10028","display_name":"Topic Modeling","score":0.7851999998092651,"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/T10028","display_name":"Topic Modeling","score":0.7851999998092651,"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/T10260","display_name":"Software Engineering Research","score":0.16220000386238098,"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/T10533","display_name":"Teaching and Learning Programming","score":0.009200000204145908,"subfield":{"id":"https://openalex.org/subfields/1706","display_name":"Computer Science Applications"},"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/llm-post-training","display_name":"LLM post-training","score":0.9380000233650208},{"id":"https://openalex.org/keywords/fine-tuning","display_name":"fine-tuning","score":0.7179999947547913},{"id":"https://openalex.org/keywords/dataset-selection","display_name":"data selection","score":0.6859999895095825},{"id":"https://openalex.org/keywords/code-generation","display_name":"code generation","score":0.7540000081062317},{"id":"https://openalex.org/keywords/instruction-tuned-models","display_name":"instruction-tuned models","score":0.7459999918937683},{"id":"https://openalex.org/keywords/less-is-more","display_name":"less is more","score":0.75}],"concepts":[{"id":"https://openalex.org/C2776760102","wikidata":"https://www.wikidata.org/wiki/Q5139990","display_name":"Code (set theory)","level":3,"score":0.734000027179718},{"id":"https://openalex.org/C2780451532","wikidata":"https://www.wikidata.org/wiki/Q759676","display_name":"Task (project management)","level":2,"score":0.6904000043869019},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6474999785423279},{"id":"https://openalex.org/C81917197","wikidata":"https://www.wikidata.org/wiki/Q628760","display_name":"Selection (genetic algorithm)","level":2,"score":0.5131000280380249},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.48750001192092896},{"id":"https://openalex.org/C204321447","wikidata":"https://www.wikidata.org/wiki/Q30642","display_name":"Natural language processing","level":1,"score":0.4846000075340271},{"id":"https://openalex.org/C43091099","wikidata":"https://www.wikidata.org/wiki/Q1067788","display_name":"Through-the-lens metering","level":3,"score":0.44760000705718994},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.4374000132083893},{"id":"https://openalex.org/C51632099","wikidata":"https://www.wikidata.org/wiki/Q3985153","display_name":"Training set","level":2,"score":0.36980000138282776},{"id":"https://openalex.org/C2776291640","wikidata":"https://www.wikidata.org/wiki/Q2912517","display_name":"Value (mathematics)","level":2,"score":0.3659999966621399},{"id":"https://openalex.org/C93959086","wikidata":"https://www.wikidata.org/wiki/Q6888345","display_name":"Model selection","level":2,"score":0.3619999885559082},{"id":"https://openalex.org/C43126263","wikidata":"https://www.wikidata.org/wiki/Q128751","display_name":"Source code","level":2,"score":0.3224000036716461},{"id":"https://openalex.org/C124101348","wikidata":"https://www.wikidata.org/wiki/Q172491","display_name":"Data mining","level":1,"score":0.2939999997615814},{"id":"https://openalex.org/C61797465","wikidata":"https://www.wikidata.org/wiki/Q1188986","display_name":"Term (time)","level":2,"score":0.28999999165534973},{"id":"https://openalex.org/C175154964","wikidata":"https://www.wikidata.org/wiki/Q380077","display_name":"Task analysis","level":3,"score":0.2833000123500824},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.2667999863624573}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:oai:arXiv.org:2609.33845","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2609.33845","pdf_url":"https://arxiv.org/pdf/2609.33845","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":null,"license_id":null,"version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},{"id":"doi:10.48550/arxiv.2609.33845","doi":"https://doi.org/10.48550/arxiv.2609.33845","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2609.33845","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:2609.33845","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2609.33845","pdf_url":"https://arxiv.org/pdf/2609.33845","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":null,"license_id":null,"version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},"sustainable_development_goals":[],"study_designs":[],"awards":[],"funders":[],"has_content":{"grobid_xml":true,"pdf":true},"content_urls":{"pdf":"https://content.openalex.org/works/W7214717067.pdf","grobid_xml":"https://content.openalex.org/works/W7214717067.grobid-xml"},"referenced_works_count":17,"referenced_works":["https://openalex.org/W2794325560","https://openalex.org/W3134642945","https://openalex.org/W3195140297","https://openalex.org/W3198243861","https://openalex.org/W3210904070","https://openalex.org/W4402667025","https://openalex.org/W4403011998","https://openalex.org/W4409362537","https://openalex.org/W4412888419","https://openalex.org/W4414460334","https://openalex.org/W4414587831","https://openalex.org/W4414820692","https://openalex.org/W4416252396","https://openalex.org/W4416373971","https://openalex.org/W7126185466","https://openalex.org/W7133216852","https://openalex.org/W7133235763"],"related_works":["https://openalex.org/W4403011998","https://openalex.org/W4414820692","https://openalex.org/W4387221974","https://openalex.org/W4403593660","https://openalex.org/W4404783276","https://openalex.org/W4403585905","https://openalex.org/W4414587831","https://openalex.org/W6910919836","https://openalex.org/W4389519352","https://openalex.org/W4392488453"],"abstract_inverted_index":{"Evaluating":[0],"code":[1,12,57,90,192],"data":[2,13,21,193],"as":[3],"a":[4,42,79,181],"single":[5],"corpus":[6,58],"can":[7,107],"obscure":[8],"which":[9,15,199,204],"types":[10],"of":[11,28,67,173,191],"benefit":[14],"models":[16,84],"and":[17,31,89,113,125,137,166,186,206],"downstream":[18],"tasks.":[19],"Effective":[20],"selection":[22],"requires":[23],"understanding":[24],"both":[25,162],"the":[26,64,104,122,158,174,189],"benefits":[27,34,140],"individual":[29,76,143],"categories":[30,37,61,77,144,200],"whether":[32,139],"these":[33,47,135],"persist":[35,145],"when":[36],"are":[38,201],"combined.":[39],"We":[40,52,128],"introduce":[41],"decompositional":[43],"lens":[44],"for":[45,203],"studying":[46],"effects":[48],"in":[49,99,142,194],"LLM":[50],"post-training.":[51],"first":[53],"decompose":[54],"an":[55],"execution-verified":[56],"into":[59],"interpretable":[60],"based":[62],"on":[63,85,198],"computational":[65],"patterns":[66],"its":[68],"solutions.":[69],"Through":[70],"controlled":[71],"fine-tuning":[72],"experiments,":[73],"we":[74],"compare":[75],"with":[78,121],"balanced":[80],"mixture":[81],"across":[82],"instruction-tuned":[83],"question":[86],"answering,":[87],"mathematics,":[88],"generation.":[91],"The":[92,116],"resulting":[93],"response":[94],"maps":[95],"reveal":[96],"recurring":[97],"gains":[98],"average":[100],"question-answering":[101],"performance,":[102],"while":[103,169],"same":[105],"category":[106,118],"improve":[108,157],"one":[109],"model":[110,124,205],"or":[111],"task":[112,160],"degrade":[114],"another.":[115],"best-performing":[117],"also":[119],"varies":[120],"starting":[123],"target":[126,159],"task.":[127,207],"then":[129],"compose":[130],"compact":[131],"mixtures":[132,153],"guided":[133],"by":[134],"results":[136],"examine":[138],"observed":[141],"under":[146],"joint":[147],"training.":[148],"On":[149],"selected":[150],"model--task":[151],"pairs,":[152],"whose":[154],"constituents":[155],"each":[156],"outperform":[161],"their":[163],"best":[164],"constituent":[165],"full-corpus":[167],"training":[168,196],"using":[170],"roughly":[171],"10--15\\%":[172],"full":[175],"corpus.":[176],"These":[177],"exploratory":[178],"findings":[179],"illustrate":[180],"\\emph{less":[182],"is":[183],"more}":[184],"pattern":[185],"highlight":[187],"how":[188],"value":[190],"post":[195],"depends":[197],"combined":[202]},"counts_by_year":[],"updated_date":"2026-11-07T11:25:07.637180","created_date":"2026-09-30T00:00:00"}