{"id":"https://openalex.org/W7224780547","doi":"https://doi.org/10.48550/arxiv.2610.39380","title":"Learning to Formulate: Agentic Hardware-Software Co-Design of AI Infrastructure through Progressive Optimization and Verification","display_name":"Learning to Formulate: Agentic Hardware-Software Co-Design of AI Infrastructure through Progressive Optimization and Verification","publication_year":2026,"publication_date":"2026-10-29","ids":{"openalex":"https://openalex.org/W7224780547","doi":"https://doi.org/10.48550/arxiv.2610.39380"},"language":"en","primary_location":{"id":"pmh:oai:arXiv.org:2610.39380","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2610.39380","pdf_url":"https://arxiv.org/pdf/2610.39380","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","indexed_in":["arxiv","datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://arxiv.org/pdf/2610.39380","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5160228928","display_name":"Yi Ding","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Ding, Yi","raw_affiliation_strings":["Shenzhen Loop Area Institute (SLAI)"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Shenzhen Loop Area Institute (SLAI)","institution_ids":[]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5160169211","display_name":"Li Shang","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I165932596","display_name":"National University of Singapore","ror":"https://ror.org/01tgyzw49","country_code":"SG","type":"education","lineage":["https://openalex.org/I165932596"]}],"countries":["SG"],"is_corresponding":false,"raw_author_name":"Shang, Li","raw_affiliation_strings":["National University of Singapore"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"National University of Singapore","institution_ids":["https://openalex.org/I165932596"]}]}],"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/T10904","display_name":"Embedded Systems Design Techniques","score":0.5795999765396118,"subfield":{"id":"https://openalex.org/subfields/1708","display_name":"Hardware and Architecture"},"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/T10904","display_name":"Embedded Systems Design Techniques","score":0.5795999765396118,"subfield":{"id":"https://openalex.org/subfields/1708","display_name":"Hardware and Architecture"},"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/T10054","display_name":"Parallel Computing and Optimization Techniques","score":0.3831000030040741,"subfield":{"id":"https://openalex.org/subfields/1708","display_name":"Hardware and Architecture"},"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/T10036","display_name":"Advanced Neural Network Applications","score":0.006000000052154064,"subfield":{"id":"https://openalex.org/subfields/1707","display_name":"Computer Vision and Pattern Recognition"},"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/ai-infrastructure","display_name":"AI infrastructure","score":0.6240000128746033},{"id":"https://openalex.org/keywords/hardware-software-co-design","display_name":"hardware-software co-design","score":0.9750000238418579},{"id":"https://openalex.org/keywords/language-model-agents","display_name":"language model agents","score":0.7300000190734863},{"id":"https://openalex.org/keywords/mathematical-formulation","display_name":"mathematical formulation","score":0.4880000054836273},{"id":"https://openalex.org/keywords/optimization","display_name":"optimization","score":0.30799999833106995},{"id":"https://openalex.org/keywords/verification","display_name":"verification","score":0.8029999732971191},{"id":"https://openalex.org/keywords/wall-relaxation","display_name":"wall relaxation","score":0.5460000038146973},{"id":"https://openalex.org/keywords/inference-acceleration","display_name":"inference acceleration","score":0.3720000088214874},{"id":"https://openalex.org/keywords/gemm","display_name":"GEMM","score":0.5400000214576721},{"id":"https://openalex.org/keywords/qwen","display_name":"Qwen","score":0.45500001311302185}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7961999773979187},{"id":"https://openalex.org/C2778770139","wikidata":"https://www.wikidata.org/wiki/Q1966904","display_name":"Solver","level":2,"score":0.5935999751091003},{"id":"https://openalex.org/C108710211","wikidata":"https://www.wikidata.org/wiki/Q11538","display_name":"Mathematical proof","level":2,"score":0.5788999795913696},{"id":"https://openalex.org/C2777904410","wikidata":"https://www.wikidata.org/wiki/Q7397","display_name":"Software","level":2,"score":0.5654000043869019},{"id":"https://openalex.org/C206588197","wikidata":"https://www.wikidata.org/wiki/Q846574","display_name":"Reuse","level":2,"score":0.5325999855995178},{"id":"https://openalex.org/C45374587","wikidata":"https://www.wikidata.org/wiki/Q12525525","display_name":"Computation","level":2,"score":0.5029000043869019},{"id":"https://openalex.org/C132525143","wikidata":"https://www.wikidata.org/wiki/Q141488","display_name":"Graph","level":2,"score":0.3874000012874603},{"id":"https://openalex.org/C80444323","wikidata":"https://www.wikidata.org/wiki/Q2878974","display_name":"Theoretical computer science","level":1,"score":0.37689998745918274},{"id":"https://openalex.org/C3019612716","wikidata":"https://www.wikidata.org/wiki/Q730920","display_name":"Problem solver","level":2,"score":0.36390000581741333},{"id":"https://openalex.org/C2778400913","wikidata":"https://www.wikidata.org/wiki/Q1639024","display_name":"Mathematical software","level":3,"score":0.3386000096797943},{"id":"https://openalex.org/C115903868","wikidata":"https://www.wikidata.org/wiki/Q80993","display_name":"Software engineering","level":1,"score":0.32919999957084656},{"id":"https://openalex.org/C2776029896","wikidata":"https://www.wikidata.org/wiki/Q3935810","display_name":"Relaxation (psychology)","level":2,"score":0.3264000117778778},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.32199999690055847},{"id":"https://openalex.org/C165005293","wikidata":"https://www.wikidata.org/wiki/Q1074500","display_name":"Chip","level":2,"score":0.3206000030040741},{"id":"https://openalex.org/C199360897","wikidata":"https://www.wikidata.org/wiki/Q9143","display_name":"Programming language","level":1,"score":0.31790000200271606},{"id":"https://openalex.org/C189950617","wikidata":"https://www.wikidata.org/wiki/Q937228","display_name":"Property (philosophy)","level":2,"score":0.3098999857902527},{"id":"https://openalex.org/C113775141","wikidata":"https://www.wikidata.org/wiki/Q428691","display_name":"Computer engineering","level":1,"score":0.3093000054359436},{"id":"https://openalex.org/C33054407","wikidata":"https://www.wikidata.org/wiki/Q6504747","display_name":"Software verification","level":5,"score":0.30660000443458557},{"id":"https://openalex.org/C164155591","wikidata":"https://www.wikidata.org/wiki/Q2067766","display_name":"Satisfiability modulo theories","level":2,"score":0.30390000343322754},{"id":"https://openalex.org/C82714645","wikidata":"https://www.wikidata.org/wiki/Q438331","display_name":"Template","level":2,"score":0.29760000109672546},{"id":"https://openalex.org/C67953723","wikidata":"https://www.wikidata.org/wiki/Q192525","display_name":"Workstation","level":2,"score":0.2971000075340271},{"id":"https://openalex.org/C120314980","wikidata":"https://www.wikidata.org/wiki/Q180634","display_name":"Distributed computing","level":1,"score":0.28929999470710754},{"id":"https://openalex.org/C2778692605","wikidata":"https://www.wikidata.org/wiki/Q4041866","display_name":"Joins","level":2,"score":0.27070000767707825},{"id":"https://openalex.org/C149091818","wikidata":"https://www.wikidata.org/wiki/Q2429814","display_name":"Software system","level":3,"score":0.2694999873638153},{"id":"https://openalex.org/C137836250","wikidata":"https://www.wikidata.org/wiki/Q984063","display_name":"Optimization problem","level":2,"score":0.25859999656677246}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:oai:arXiv.org:2610.39380","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2610.39380","pdf_url":"https://arxiv.org/pdf/2610.39380","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.2610.39380","doi":"https://doi.org/10.48550/arxiv.2610.39380","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2610.39380","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:2610.39380","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2610.39380","pdf_url":"https://arxiv.org/pdf/2610.39380","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/W7224780547.pdf","grobid_xml":"https://content.openalex.org/works/W7224780547.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":["https://openalex.org/W7166074791","https://openalex.org/W7160874444","https://openalex.org/W7162628148","https://openalex.org/W4415744094","https://openalex.org/W7110812152","https://openalex.org/W7109077862","https://openalex.org/W4416931995","https://openalex.org/W7203615598","https://openalex.org/W7222835251","https://openalex.org/W7172188853"],"abstract_inverted_index":{"AI":[0,10],"infrastructure,":[1],"the":[2,18,34,47,83,88,108,114,118,130,140,144,149,169,189,197],"chips":[3],"and":[4,8,36,96,98,103],"software":[5,38,202],"that":[6,32,69,81,147],"train":[7],"serve":[9],"models,":[11],"is":[12,72,122],"designed":[13],"in":[14],"layers.":[15],"Layers":[16],"make":[17,54],"system":[19],"manageable,":[20],"but":[21,61],"they":[22,134],"also":[23],"fix":[24],"which":[25,133],"choices":[26],"an":[27,62],"optimizer":[28],"can":[29],"consider.":[30],"Gains":[31],"need":[33],"chip":[35,141],"its":[37,101],"to":[39,57,194],"change":[40],"together":[41],"are":[42,135],"excluded":[43],"before":[44],"optimization":[45],"begins:":[46],"formulation":[48],"leaves":[49],"them":[50],"out.":[51],"Language-model":[52],"agents":[53],"it":[55],"possible":[56],"revisit":[58],"these":[59],"boundaries,":[60],"agent's":[63],"proposal":[64],"cannot":[65],"show":[66],"by":[67,111,142],"itself":[68],"a":[70,79,90,123,125,154,181,205],"design":[71,187],"correct":[73],"or":[74],"optimal.":[75],"We":[76],"present":[77],"Formulaic,":[78],"flow":[80],"divides":[82],"work.":[84],"The":[85,105],"agent":[86,106,170],"writes":[87],"formulation,":[89],"mathematical":[91],"solver":[92],"optimizes":[93],"within":[94],"it,":[95],"verification":[97],"measurement":[99],"check":[100],"assumptions":[102],"results.":[104],"refines":[107],"problem":[109],"step":[110,121],"step,":[112,156],"from":[113],"computation":[115],"graph":[116],"toward":[117],"chip.":[119],"Each":[120],"template:":[124],"family":[126],"of":[127,192,196],"programs":[128],"with":[129],"conditions":[131],"under":[132],"correct.":[136],"Wall":[137],"relaxation":[138],"searches":[139],"pricing":[143],"hardware":[145],"limits":[146],"bind":[148],"software.":[150],"This":[151],"paper":[152],"takes":[153],"first":[155],"through":[157],"separate":[158],"studies":[159],"rather":[160],"than":[161,178],"one":[162],"end-to-end":[163],"run.":[164],"On":[165,204],"NVIDIA":[166],"GB10,":[167],"templates":[168],"wrote":[171],"run":[172],"complete":[173],"Qwen3-8B":[174],"layers":[175],"$1.075\\times$":[176],"faster":[177],"SGLang.":[179],"In":[180],"simulated":[182],"Ventus":[183],"GPU":[184],"study,":[185],"joint":[186],"cuts":[188],"worst-scenario":[190],"cycles":[191],"Qwen2.5-0.5B":[193],"0.428":[195],"reference,":[198],"against":[199],"0.934":[200],"for":[201],"alone.":[203],"five-step":[206],"GEMM":[207],"refinement":[208],"chain,":[209],"reusing":[210],"proofs":[211],"makes":[212],"checking":[213],"$2.51\\times$":[214],"faster.":[215]},"counts_by_year":[],"updated_date":"2026-11-06T12:05:46.655746","created_date":"2026-10-31T00:00:00"}