{"id":"https://openalex.org/W7204246716","doi":"https://doi.org/10.48550/arxiv.2608.24571","title":"Joint Optimization of Tool Creation and Use for Large Language Model Agents","display_name":"Joint Optimization of Tool Creation and Use for Large Language Model Agents","publication_year":2026,"publication_date":"2026-08-25","ids":{"openalex":"https://openalex.org/W7204246716","doi":"https://doi.org/10.48550/arxiv.2608.24571"},"language":"en","primary_location":{"id":"pmh:oai:arXiv.org:2608.24571","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2608.24571","pdf_url":"https://arxiv.org/pdf/2608.24571","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","indexed_in":["arxiv","datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://arxiv.org/pdf/2608.24571","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5028150164","display_name":"Zhi Rui Tam","orcid":"https://orcid.org/0000-0001-9968-2416","observed_orcids":["https://orcid.org/0000-0001-9968-2416"]},"institutions":[{"id":"https://openalex.org/I16733864","display_name":"National Taiwan University","ror":"https://ror.org/05bqach95","country_code":"TW","type":"education","lineage":["https://openalex.org/I16733864"]}],"countries":["TW"],"is_corresponding":false,"raw_author_name":"Tam, Zhi Rui","raw_affiliation_strings":["Appier AI Research ,","National Taiwan University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Appier AI Research ,","institution_ids":[]},{"raw_affiliation_string":"National Taiwan University","institution_ids":["https://openalex.org/I16733864"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5101379633","display_name":"Chieh-Yen Lin","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Lin, Chieh-Yen","raw_affiliation_strings":["Appier AI Research ,"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Appier AI Research ,","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5126220772","display_name":"Yun-Nung Chen","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I16733864","display_name":"National Taiwan University","ror":"https://ror.org/05bqach95","country_code":"TW","type":"education","lineage":["https://openalex.org/I16733864"]}],"countries":["TW"],"is_corresponding":false,"raw_author_name":"Chen, Yun-Nung","raw_affiliation_strings":["National Taiwan University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"National Taiwan University","institution_ids":["https://openalex.org/I16733864"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5126225448","display_name":"Shao-Hua Sun","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I16733864","display_name":"National Taiwan University","ror":"https://ror.org/05bqach95","country_code":"TW","type":"education","lineage":["https://openalex.org/I16733864"]}],"countries":["TW"],"is_corresponding":false,"raw_author_name":"Sun, Shao-Hua","raw_affiliation_strings":["Appier AI Research ,","National Taiwan University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Appier AI Research ,","institution_ids":[]},{"raw_affiliation_string":"National Taiwan University","institution_ids":["https://openalex.org/I16733864"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5139212885","display_name":"Hung-yi Lee","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I16733864","display_name":"National Taiwan University","ror":"https://ror.org/05bqach95","country_code":"TW","type":"education","lineage":["https://openalex.org/I16733864"]}],"countries":["TW"],"is_corresponding":false,"raw_author_name":"Lee, Hung-yi","raw_affiliation_strings":["National Taiwan University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"National Taiwan University","institution_ids":["https://openalex.org/I16733864"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":1,"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":0.27320000529289246,"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":0.27320000529289246,"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/T10906","display_name":"AI-based Problem Solving and Planning","score":0.23899999260902405,"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":0.1476999968290329,"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/tool-use","display_name":"tool use","score":0.9300000071525574},{"id":"https://openalex.org/keywords/llm-based-agents","display_name":"LLM agents","score":0.8949999809265137},{"id":"https://openalex.org/keywords/reinforcement-learning","display_name":"reinforcement learning","score":0.546999990940094},{"id":"https://openalex.org/keywords/multi-task-learning","display_name":"multi-task learning","score":0.6019999980926514},{"id":"https://openalex.org/keywords/procedural-reasoning","display_name":"procedural reasoning","score":0.7429999709129333},{"id":"https://openalex.org/keywords/reward-function-design","display_name":"reward function design","score":0.32899999618530273},{"id":"https://openalex.org/keywords/qwen3","display_name":"Qwen3","score":0.5509999990463257},{"id":"https://openalex.org/keywords/gqa","display_name":"GQA","score":0.5130000114440918},{"id":"https://openalex.org/keywords/llm-as-a-judge","display_name":"LLM-as-a-judge","score":0.3930000066757202}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7429999709129333},{"id":"https://openalex.org/C2780451532","wikidata":"https://www.wikidata.org/wiki/Q759676","display_name":"Task (project management)","level":2,"score":0.6916000247001648},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.6050000190734863},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.5889000296592712},{"id":"https://openalex.org/C148220186","wikidata":"https://www.wikidata.org/wiki/Q7111912","display_name":"Outcome (game theory)","level":2,"score":0.5756999850273132},{"id":"https://openalex.org/C97541855","wikidata":"https://www.wikidata.org/wiki/Q830687","display_name":"Reinforcement learning","level":2,"score":0.5514000058174133},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.5461000204086304},{"id":"https://openalex.org/C190839683","wikidata":"https://www.wikidata.org/wiki/Q2448197","display_name":"Train","level":2,"score":0.47530001401901245},{"id":"https://openalex.org/C22367795","wikidata":"https://www.wikidata.org/wiki/Q7625208","display_name":"Structured prediction","level":2,"score":0.44190001487731934},{"id":"https://openalex.org/C175154964","wikidata":"https://www.wikidata.org/wiki/Q380077","display_name":"Task analysis","level":3,"score":0.42829999327659607},{"id":"https://openalex.org/C18555067","wikidata":"https://www.wikidata.org/wiki/Q8375051","display_name":"Joint (building)","level":2,"score":0.4052000045776367},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.35249999165534973},{"id":"https://openalex.org/C48677424","wikidata":"https://www.wikidata.org/wiki/Q6888088","display_name":"Mode (computer interface)","level":2,"score":0.3506999909877777},{"id":"https://openalex.org/C107457646","wikidata":"https://www.wikidata.org/wiki/Q207434","display_name":"Human\u2013computer interaction","level":1,"score":0.3476000130176544},{"id":"https://openalex.org/C34388435","wikidata":"https://www.wikidata.org/wiki/Q2267362","display_name":"Bounded function","level":2,"score":0.33959999680519104},{"id":"https://openalex.org/C2779843651","wikidata":"https://www.wikidata.org/wiki/Q7390335","display_name":"SIGNAL (programming language)","level":2,"score":0.3278999924659729},{"id":"https://openalex.org/C2779982483","wikidata":"https://www.wikidata.org/wiki/Q6094420","display_name":"Iterative refinement","level":2,"score":0.3127000033855438},{"id":"https://openalex.org/C204321447","wikidata":"https://www.wikidata.org/wiki/Q30642","display_name":"Natural language processing","level":1,"score":0.2667999863624573},{"id":"https://openalex.org/C2777508537","wikidata":"https://www.wikidata.org/wiki/Q7936620","display_name":"Visual reasoning","level":2,"score":0.26660001277923584},{"id":"https://openalex.org/C98045186","wikidata":"https://www.wikidata.org/wiki/Q205663","display_name":"Process (computing)","level":2,"score":0.26440000534057617},{"id":"https://openalex.org/C59656382","wikidata":"https://www.wikidata.org/wiki/Q191536","display_name":"Conjunction (astronomy)","level":2,"score":0.2547000050544739},{"id":"https://openalex.org/C51632099","wikidata":"https://www.wikidata.org/wiki/Q3985153","display_name":"Training set","level":2,"score":0.25029999017715454}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:oai:arXiv.org:2608.24571","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2608.24571","pdf_url":"https://arxiv.org/pdf/2608.24571","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.2608.24571","doi":"https://doi.org/10.48550/arxiv.2608.24571","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2608.24571","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:2608.24571","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2608.24571","pdf_url":"https://arxiv.org/pdf/2608.24571","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":[{"id":"https://openalex.org/F4320331164","display_name":"National Science and Technology Council","ror":"https://ror.org/00wnb9798"}],"has_content":{"grobid_xml":true,"pdf":true},"content_urls":{"pdf":"https://content.openalex.org/works/W7204246716.pdf","grobid_xml":"https://content.openalex.org/works/W7204246716.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":["https://openalex.org/W4415317186","https://openalex.org/W4416393082","https://openalex.org/W7126215865","https://openalex.org/W4416758252","https://openalex.org/W4399115751","https://openalex.org/W4404782859","https://openalex.org/W4409657210","https://openalex.org/W4387294587","https://openalex.org/W7119481089","https://openalex.org/W4389520009"],"abstract_inverted_index":{"Tool-augmented":[0],"language":[1],"models":[2,185],"are":[3,47],"bounded":[4],"by":[5,17,182],"the":[6,26,34,43,142,168,188],"APIs":[7],"humans":[8],"bothered":[9],"to":[10],"write;":[11],"existing":[12],"tool-creation":[13],"systems":[14],"patch":[15],"this":[16],"prompting":[18],"a":[19,30,60,73,80,84,87,91,95,99],"frozen":[20],"LLM":[21],"at":[22],"inference":[23],"time,":[24],"leaving":[25],"model":[27],"that":[28,36,42,64],"writes":[29],"tool":[31,67,70,85,97],"decoupled":[32],"from":[33,86],"one":[35],"uses":[37],"it,":[38],"with":[39,125,132],"no":[40],"signal":[41],"schemas":[44,48],"it":[45,49],"produces":[46],"can":[50],"invoke.":[51],"We":[52],"propose":[53],"SMITH":[54,126],"(Schema-grounded":[55],"Multi-task":[56],"Iterative":[57],"Tool":[58],"Honing),":[59],"reinforcement":[61],"learning":[62],"framework":[63],"jointly":[65],"trains":[66],"creation":[68],"and":[69,109,148,161,192],"use":[71,92],"inside":[72],"single":[74],"policy.":[75],"Each":[76],"rollout":[77],"is":[78],"either":[79],"build":[81],"task":[82,93],"(write":[83],"few":[88],"examples)":[89],"or":[90,176],"(invoke":[94],"pooled":[96],"on":[98,127,139,159,163],"held-out":[100,140],"question).":[101],"Three":[102],"separate":[103],"reward":[104],"axes":[105],"catch":[106],"schema,":[107],"code,":[108],"outcome":[110],"failures":[111],"independently,":[112],"so":[113],"each":[114],"failure":[115],"mode":[116],"contributes":[117],"its":[118],"own":[119],"gradient.":[120],"A":[121],"4B":[122,184],"Qwen3":[123],"trained":[124],"13":[128],"procedural":[129],"reasoning":[130,196],"tasks":[131],"exact":[133],"verifiers":[134],"reaches":[135,157],"79.8":[136],"macro-average":[137],"accuracy":[138],"tasks,":[141],"best":[143,169],"across":[144],"all":[145],"evaluated":[146],"methods":[147],"ahead":[149],"of":[150,190],"an":[151],"untrained":[152],"30B-A3B":[153],"tool-writer.":[154],"It":[155],"also":[156,186],"40.4":[158],"TabMWP-Hard":[160],"42.6":[162],"out-of-domain":[164],"GQA":[165],"(+7.6":[166],"over":[167],"same-backbone":[170],"inference-time":[171],"baseline),":[172],"without":[173],"any":[174],"visual":[175],"tabular":[177],"training":[178],"data.":[179],"Tools":[180],"written":[181],"our":[183],"lifted":[187],"performance":[189],"LFM-2.5-350M":[191],"Qwen3-30B-A3B":[193],"under":[194],"same":[195],"tasks.":[197]},"counts_by_year":[],"updated_date":"2026-10-28T10:33:24.094632","created_date":"2026-08-27T00:00:00"}