{"id":"https://openalex.org/W7214928276","doi":"https://doi.org/10.48550/arxiv.2609.37356","title":"Teaching LLMs to Generate Challenging MILP Instances via Solver Feedback","display_name":"Teaching LLMs to Generate Challenging MILP Instances via Solver Feedback","publication_year":2026,"publication_date":"2026-09-29","ids":{"openalex":"https://openalex.org/W7214928276","doi":"https://doi.org/10.48550/arxiv.2609.37356"},"language":"en","primary_location":{"id":"pmh:oai:arXiv.org:2609.37356","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2609.37356","pdf_url":"https://arxiv.org/pdf/2609.37356","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/2609.37356","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5001845207","display_name":"Jitin Singla","orcid":"https://orcid.org/0000-0001-6225-0783","observed_orcids":["https://orcid.org/0000-0001-6225-0783"]},"institutions":[{"id":"https://openalex.org/I154851008","display_name":"Indian Institute of Technology Roorkee","ror":"https://ror.org/00582g326","country_code":"IN","type":"education","lineage":["https://openalex.org/I154851008"]}],"countries":["IN"],"is_corresponding":false,"raw_author_name":"Singla, Jitin","raw_affiliation_strings":["IIT Roorkee"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"IIT Roorkee","institution_ids":["https://openalex.org/I154851008"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5055795558","display_name":"Parikshit Pareek","orcid":"https://orcid.org/0000-0003-4688-2021","observed_orcids":["https://orcid.org/0000-0003-4688-2021"]},"institutions":[{"id":"https://openalex.org/I154851008","display_name":"Indian Institute of Technology Roorkee","ror":"https://ror.org/00582g326","country_code":"IN","type":"education","lineage":["https://openalex.org/I154851008"]}],"countries":["IN"],"is_corresponding":false,"raw_author_name":"Pareek, Parikshit","raw_affiliation_strings":["IIT Roorkee"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"IIT Roorkee","institution_ids":["https://openalex.org/I154851008"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5086696547","display_name":"Pratik Jawanpuria","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I162827531","display_name":"Indian Institute of Technology Bombay","ror":"https://ror.org/02qyf5152","country_code":"IN","type":"education","lineage":["https://openalex.org/I162827531"]}],"countries":["IN"],"is_corresponding":false,"raw_author_name":"Jawanpuria, Pratik","raw_affiliation_strings":["IIT Bombay"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"IIT Bombay","institution_ids":["https://openalex.org/I162827531"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5103677687","display_name":"Parag Singla","orcid":"https://orcid.org/0009-0000-9190-9794","observed_orcids":["https://orcid.org/0009-0000-9190-9794"]},"institutions":[{"id":"https://openalex.org/I68891433","display_name":"Indian Institute of Technology Delhi","ror":"https://ror.org/049tgcd06","country_code":"IN","type":"education","lineage":["https://openalex.org/I68891433"]}],"countries":["IN"],"is_corresponding":false,"raw_author_name":"Singla, Parag","raw_affiliation_strings":["IIT Delhi"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"IIT Delhi","institution_ids":["https://openalex.org/I68891433"]}]}],"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/T10963","display_name":"Advanced Optimization Algorithms Research","score":0.5054000020027161,"subfield":{"id":"https://openalex.org/subfields/2612","display_name":"Numerical Analysis"},"field":{"id":"https://openalex.org/fields/26","display_name":"Mathematics"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},"topics":[{"id":"https://openalex.org/T10963","display_name":"Advanced Optimization Algorithms Research","score":0.5054000020027161,"subfield":{"id":"https://openalex.org/subfields/2612","display_name":"Numerical Analysis"},"field":{"id":"https://openalex.org/fields/26","display_name":"Mathematics"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},{"id":"https://openalex.org/T10028","display_name":"Topic Modeling","score":0.08340000361204147,"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/T10100","display_name":"Metaheuristic Optimization Algorithms Research","score":0.06610000133514404,"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/problem-difficulty","display_name":"problem difficulty","score":0.4830000102519989},{"id":"https://openalex.org/keywords/capacitated-facility-location","display_name":"capacitated facility location","score":0.7419999837875366},{"id":"https://openalex.org/keywords/max-cut","display_name":"MaxCut","score":0.9900000095367432},{"id":"https://openalex.org/keywords/learning-based-optimization","display_name":"learning-based optimization","score":0.5429999828338623}],"concepts":[{"id":"https://openalex.org/C2778770139","wikidata":"https://www.wikidata.org/wiki/Q1966904","display_name":"Solver","level":2,"score":0.7730000019073486},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7019000053405762},{"id":"https://openalex.org/C153083717","wikidata":"https://www.wikidata.org/wiki/Q6535263","display_name":"Leverage (statistics)","level":2,"score":0.7013999819755554},{"id":"https://openalex.org/C126255220","wikidata":"https://www.wikidata.org/wiki/Q141495","display_name":"Mathematical optimization","level":1,"score":0.586899995803833},{"id":"https://openalex.org/C86251818","wikidata":"https://www.wikidata.org/wiki/Q816754","display_name":"Benchmarking","level":2,"score":0.534500002861023},{"id":"https://openalex.org/C26517878","wikidata":"https://www.wikidata.org/wiki/Q228039","display_name":"Key (lock)","level":2,"score":0.5063999891281128},{"id":"https://openalex.org/C2776029896","wikidata":"https://www.wikidata.org/wiki/Q3935810","display_name":"Relaxation (psychology)","level":2,"score":0.3714999854564667},{"id":"https://openalex.org/C204323151","wikidata":"https://www.wikidata.org/wiki/Q905424","display_name":"Range (aeronautics)","level":2,"score":0.3359000086784363},{"id":"https://openalex.org/C113174947","wikidata":"https://www.wikidata.org/wiki/Q2859736","display_name":"Tree (set theory)","level":2,"score":0.3319000005722046},{"id":"https://openalex.org/C137836250","wikidata":"https://www.wikidata.org/wiki/Q984063","display_name":"Optimization problem","level":2,"score":0.3237000107765198},{"id":"https://openalex.org/C56086750","wikidata":"https://www.wikidata.org/wiki/Q6042592","display_name":"Integer programming","level":2,"score":0.31859999895095825},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.30230000615119934},{"id":"https://openalex.org/C97541855","wikidata":"https://www.wikidata.org/wiki/Q830687","display_name":"Reinforcement learning","level":2,"score":0.29600000381469727},{"id":"https://openalex.org/C25360446","wikidata":"https://www.wikidata.org/wiki/Q1512771","display_name":"Linear programming relaxation","level":3,"score":0.2921000123023987},{"id":"https://openalex.org/C173801870","wikidata":"https://www.wikidata.org/wiki/Q201413","display_name":"Heuristic","level":2,"score":0.2897999882698059},{"id":"https://openalex.org/C48372109","wikidata":"https://www.wikidata.org/wiki/Q3913","display_name":"Binary number","level":2,"score":0.2808000147342682},{"id":"https://openalex.org/C55660270","wikidata":"https://www.wikidata.org/wiki/Q5164377","display_name":"Constrained optimization","level":2,"score":0.2775999903678894},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.27730000019073486},{"id":"https://openalex.org/C80444323","wikidata":"https://www.wikidata.org/wiki/Q2878974","display_name":"Theoretical computer science","level":1,"score":0.266400009393692}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:oai:arXiv.org:2609.37356","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2609.37356","pdf_url":"https://arxiv.org/pdf/2609.37356","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.2609.37356","doi":"https://doi.org/10.48550/arxiv.2609.37356","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2609.37356","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:2609.37356","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2609.37356","pdf_url":"https://arxiv.org/pdf/2609.37356","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":[{"id":"https://openalex.org/F4320306227","display_name":"Arthritis National Research Foundation","ror":"https://ror.org/03abbex33"}],"has_content":{"grobid_xml":true,"pdf":true},"content_urls":{"pdf":"https://content.openalex.org/works/W7214928276.pdf","grobid_xml":"https://content.openalex.org/works/W7214928276.grobid-xml"},"referenced_works_count":14,"referenced_works":["https://openalex.org/W1974709582","https://openalex.org/W2014328192","https://openalex.org/W2071622945","https://openalex.org/W2189345514","https://openalex.org/W2900896126","https://openalex.org/W3085454911","https://openalex.org/W3193251185","https://openalex.org/W4391631327","https://openalex.org/W4399389097","https://openalex.org/W4399598128","https://openalex.org/W4410192806","https://openalex.org/W4414939355","https://openalex.org/W4417299665","https://openalex.org/W7167604932"],"related_works":["https://openalex.org/W4389048934","https://openalex.org/W4387560731","https://openalex.org/W6948060991","https://openalex.org/W7133203436","https://openalex.org/W7131831054","https://openalex.org/W4416245577","https://openalex.org/W7166864383","https://openalex.org/W7167264377","https://openalex.org/W4401201897","https://openalex.org/W4403579809"],"abstract_inverted_index":{"Generating":[0],"optimization":[1,18,66,254],"instances":[2,26,121,135],"that":[3,64,86,234,241],"are":[4,136],"both":[5,88],"feasible":[6,155],"and":[7,15,79,91,101,122,127,141,145,151,168,178,187,225,227,261],"computationally":[8],"challenging":[9,60,157],"is":[10,56,108],"crucial":[11],"for":[12,232],"benchmarking":[13],"solvers":[14],"training":[16,133],"learning-based":[17],"algorithms.":[19],"Existing":[20],"non-LLM":[21],"generators":[22,39],"rely":[23],"on":[24,192,223,264],"seed":[25,131],"or":[27,132],"parameter":[28],"tuning,":[29],"resulting":[30],"in":[31,245],"high":[32],"test-time":[33],"computational":[34],"cost,":[35],"while":[36,198],"existing":[37],"LLM":[38,116],"lack":[40],"explicit":[41],"hardness":[42,93],"measures.":[43],"Recent":[44],"reinforcement":[45],"learning":[46],"methods":[47],"with":[48,58,143],"verifier":[49],"feedback":[50],"evaluate":[51],"only":[52],"binary":[53],"correctness,":[54],"which":[55,153],"misaligned":[57],"generating":[59],"problems.":[61],"We":[62,138,256],"note":[63],"an":[65,115],"solver":[67,124,230],"reports":[68],"the":[69,89,92,123,189,218],"cost":[70],"of":[71,76,94,217],"solving":[72],"at":[73],"several":[74],"stages":[75],"its":[77,184],"pipeline,":[78],"leverage":[80],"this":[81],"to":[82,205,251],"design":[83],"a":[84,109,146,210],"reward":[85],"scores":[87],"solvability":[90],"generated":[95],"problems,":[96],"measured":[97],"by":[98,176,181,195,203],"branch-and-bound":[99],"nodes":[100,175,202],"post-cut":[102,179],"relaxation":[103],"gaps.":[104],"Our":[105],"key":[106],"idea":[107],"challenger-solver":[110],"asymmetric":[111],"self-play":[112,246],"approach,":[113],"where":[114],"challenger":[117],"generates":[118],"progressively":[119],"harder":[120],"verifies":[125],"feasibility":[126,190],"hardness,":[128],"so":[129],"no":[130],"MILP":[134,158],"required.":[137],"fine-tune":[139],"Gemma-4-12B":[140],"Qwen3.5-4B":[142],"GRPO":[144],"size":[147],"curriculum":[148],"into":[149],"OptiScribe-12B":[150,170],"OptiScribe-4B,":[152],"generate":[154,252],"yet":[156],"problems":[159,208],"from":[160],"natural":[161],"language":[162],"instructions.":[163],"On":[164],"capacitated":[165],"facility":[166,193],"location":[167,194],"max-cut,":[169],"raises":[171,200],"median":[172,201],"SCIP":[173],"search":[174],"1.7-5x":[177],"gaps":[180],"1.1-1.7x":[182],"over":[183],"base":[185],"model":[186],"improves":[188],"rate":[191],"9-19":[196],"points,":[197],"OptiScribe-4B":[199],"up":[204],"15.6x.":[206],"The":[207],"cover":[209],"wider":[211],"difficulty":[212],"range":[213],"than":[214],"public":[215,235],"benchmarks":[216],"same":[219],"size,":[220],"follow":[221],"instructions":[222],"density":[224],"difficulty,":[226],"can":[228,248],"tune":[229],"settings":[231],"families":[233],"libraries":[236],"lack.":[237],"These":[238],"results":[239],"indicate":[240],"optimization-specific":[242],"rewards,":[243],"used":[244],"mode,":[247],"teach":[249],"LLMs":[250],"high-difficulty":[253],"benchmarks.":[255],"will":[257],"release":[258],"our":[259],"code":[260],"models":[262],"publicly":[263],"acceptance.":[265]},"counts_by_year":[],"updated_date":"2026-11-07T11:25:07.637180","created_date":"2026-10-01T00:00:00"}