{"id":"https://openalex.org/W7165946694","doi":"https://doi.org/10.48550/arxiv.2606.25432","title":"Brevity is the Soul of Inference Efficiency: Inducing Concision in VLMs via Data Curation","display_name":"Brevity is the Soul of Inference Efficiency: Inducing Concision in VLMs via Data Curation","publication_year":2026,"publication_date":"2026-06-24","ids":{"openalex":"https://openalex.org/W7165946694","doi":"https://doi.org/10.48550/arxiv.2606.25432"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2606.25432","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2606.25432","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,"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"},"type":"preprint","indexed_in":["datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://doi.org/10.48550/arxiv.2606.25432","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5139376211","display_name":"DatologyAI","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"DatologyAI","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5139320297","display_name":":","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":":","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5036343150","display_name":"Matthew L. Leavitt","orcid":"https://orcid.org/0000-0003-0779-108X"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Leavitt, Matthew L.","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5139323031","display_name":"Siddharth Joshi","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Joshi, Siddharth","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5062209636","display_name":"Haoli Yin","orcid":"https://orcid.org/0000-0001-7842-3474"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Yin, Haoli","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5094095396","display_name":"Rishabh Adiga","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Adiga, Rishabh","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5139341975","display_name":"Haakon Mongstad","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Mongstad, Haakon","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5139344155","display_name":"Alvin Deng","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Deng, Alvin","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5135921997","display_name":"David Schwab","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Schwab, David","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5139378493","display_name":"Bogdan Gaza","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Gaza, Bogdan","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5139323462","display_name":"Ari Morcos","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Morcos, Ari","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":0,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":null,"has_fulltext":false,"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/T11714","display_name":"Multimodal Machine Learning Applications","score":0.2150000035762787,"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"}},"topics":[{"id":"https://openalex.org/T11714","display_name":"Multimodal Machine Learning Applications","score":0.2150000035762787,"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"}},{"id":"https://openalex.org/T10775","display_name":"Generative Adversarial Networks and Image Synthesis","score":0.07569999992847443,"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"}},{"id":"https://openalex.org/T10028","display_name":"Topic Modeling","score":0.05040000006556511,"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/inference","display_name":"Inference","score":0.7128999829292297},{"id":"https://openalex.org/keywords/pipeline","display_name":"Pipeline (software)","score":0.6334999799728394},{"id":"https://openalex.org/keywords/set","display_name":"Set (abstract data type)","score":0.593999981880188},{"id":"https://openalex.org/keywords/data-set","display_name":"Data set","score":0.4787999987602234},{"id":"https://openalex.org/keywords/scale","display_name":"Scale (ratio)","score":0.47769999504089355},{"id":"https://openalex.org/keywords/equating","display_name":"Equating","score":0.46720001101493835},{"id":"https://openalex.org/keywords/security-token","display_name":"Security token","score":0.45719999074935913},{"id":"https://openalex.org/keywords/baseline","display_name":"Baseline (sea)","score":0.44589999318122864},{"id":"https://openalex.org/keywords/data-curation","display_name":"Data curation","score":0.4442000091075897}],"concepts":[{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.7128999829292297},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6740999817848206},{"id":"https://openalex.org/C43521106","wikidata":"https://www.wikidata.org/wiki/Q2165493","display_name":"Pipeline (software)","level":2,"score":0.6334999799728394},{"id":"https://openalex.org/C177264268","wikidata":"https://www.wikidata.org/wiki/Q1514741","display_name":"Set (abstract data type)","level":2,"score":0.593999981880188},{"id":"https://openalex.org/C58489278","wikidata":"https://www.wikidata.org/wiki/Q1172284","display_name":"Data set","level":2,"score":0.4787999987602234},{"id":"https://openalex.org/C2778755073","wikidata":"https://www.wikidata.org/wiki/Q10858537","display_name":"Scale (ratio)","level":2,"score":0.47769999504089355},{"id":"https://openalex.org/C106347477","wikidata":"https://www.wikidata.org/wiki/Q5384228","display_name":"Equating","level":3,"score":0.46720001101493835},{"id":"https://openalex.org/C48145219","wikidata":"https://www.wikidata.org/wiki/Q1335365","display_name":"Security token","level":2,"score":0.45719999074935913},{"id":"https://openalex.org/C12725497","wikidata":"https://www.wikidata.org/wiki/Q810247","display_name":"Baseline (sea)","level":2,"score":0.44589999318122864},{"id":"https://openalex.org/C91632574","wikidata":"https://www.wikidata.org/wiki/Q15088675","display_name":"Data curation","level":2,"score":0.4442000091075897},{"id":"https://openalex.org/C67186912","wikidata":"https://www.wikidata.org/wiki/Q367664","display_name":"Data modeling","level":2,"score":0.44369998574256897},{"id":"https://openalex.org/C129916263","wikidata":"https://www.wikidata.org/wiki/Q1141183","display_name":"Backward chaining","level":4,"score":0.4422999918460846},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.4399999976158142},{"id":"https://openalex.org/C9357733","wikidata":"https://www.wikidata.org/wiki/Q6878417","display_name":"Missing data","level":2,"score":0.36059999465942383},{"id":"https://openalex.org/C149782125","wikidata":"https://www.wikidata.org/wiki/Q160039","display_name":"Econometrics","level":1,"score":0.3562999963760376},{"id":"https://openalex.org/C124101348","wikidata":"https://www.wikidata.org/wiki/Q172491","display_name":"Data mining","level":1,"score":0.34380000829696655},{"id":"https://openalex.org/C204321447","wikidata":"https://www.wikidata.org/wiki/Q30642","display_name":"Natural language processing","level":1,"score":0.34290000796318054},{"id":"https://openalex.org/C134261354","wikidata":"https://www.wikidata.org/wiki/Q938438","display_name":"Statistical inference","level":2,"score":0.3294999897480011},{"id":"https://openalex.org/C23123220","wikidata":"https://www.wikidata.org/wiki/Q816826","display_name":"Information retrieval","level":1,"score":0.31769999861717224},{"id":"https://openalex.org/C160920958","wikidata":"https://www.wikidata.org/wiki/Q7662746","display_name":"Synthetic data","level":2,"score":0.3176000118255615},{"id":"https://openalex.org/C168167062","wikidata":"https://www.wikidata.org/wiki/Q1117970","display_name":"Component (thermodynamics)","level":2,"score":0.30799999833106995},{"id":"https://openalex.org/C2778751112","wikidata":"https://www.wikidata.org/wiki/Q835016","display_name":"Window (computing)","level":2,"score":0.3068999946117401},{"id":"https://openalex.org/C167981075","wikidata":"https://www.wikidata.org/wiki/Q2667186","display_name":"Sandbox (software development)","level":2,"score":0.3043000102043152},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.2992999851703644},{"id":"https://openalex.org/C2781249084","wikidata":"https://www.wikidata.org/wiki/Q908656","display_name":"Preference","level":2,"score":0.29809999465942383},{"id":"https://openalex.org/C133462117","wikidata":"https://www.wikidata.org/wiki/Q4929239","display_name":"Data collection","level":2,"score":0.2791999876499176},{"id":"https://openalex.org/C51485801","wikidata":"https://www.wikidata.org/wiki/Q16966861","display_name":"Efficient frontier","level":3,"score":0.2770000100135803},{"id":"https://openalex.org/C105795698","wikidata":"https://www.wikidata.org/wiki/Q12483","display_name":"Statistics","level":1,"score":0.2718000113964081},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.2676999866962433},{"id":"https://openalex.org/C94375191","wikidata":"https://www.wikidata.org/wiki/Q11205","display_name":"Arithmetic","level":1,"score":0.26750001311302185},{"id":"https://openalex.org/C5655090","wikidata":"https://www.wikidata.org/wiki/Q192588","display_name":"Relational database","level":2,"score":0.26600000262260437},{"id":"https://openalex.org/C77088390","wikidata":"https://www.wikidata.org/wiki/Q8513","display_name":"Database","level":1,"score":0.2606000006198883}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2606.25432","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2606.25432","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,"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":null,"raw_source_name":null,"raw_type":"Preprint"}],"best_oa_location":{"id":"doi:10.48550/arxiv.2606.25432","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2606.25432","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,"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"},"sustainable_development_goals":[],"awards":[],"funders":[],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Inference":[0,269],"efficiency":[1,270],"is":[2,33,47,57,274,280],"typically":[3],"pursued":[4],"by":[5],"shrinking":[6],"the":[7,35,37,48,90,102,151,183,218,244,251,281],"model:":[8],"distillation,":[9],"pruning,":[10],"quantization,":[11],"and":[12,31,52,94,106,116,136,217,278],"sparse":[13],"routing":[14],"each":[15],"lower":[16,81],"per-token":[17],"cost":[18],"while":[19],"treating":[20],"token":[21],"count":[22],"as":[23,258],"fixed.":[24],"But":[25],"output":[26,125],"length":[27,126],"has":[28,79],"been":[29],"inflating,":[30],"it":[32,78,285],"precisely":[34],"component":[36],"standard":[38,103],"toolkit":[39],"leaves":[40],"untouched.":[41],"Here,":[42],"we":[43,123],"argue":[44],"that":[45,53,186,283],"brevity":[46,133,202,266,279],"missing":[49],"inference-efficiency":[50],"lever,":[51],"pretraining":[54],"data":[55,70],"curation":[56,87,261],"a":[58,64,80,112,129,146,177,259,275],"practical":[59],"way":[60],"to":[61,72,89,196,236],"pull":[62],"it:":[63],"model":[65,189,246,254],"trained":[66,97],"on":[67,98],"concise,":[68],"correct":[69,142,167,249],"learns":[71],"answer":[73],"in":[74,139,271],"fewer":[75],"tokens;":[76],"i.e.":[77],"Cost-of-Pass.":[82],"We":[83],"apply":[84],"our":[85,99],"VLM":[86],"pipeline":[88],"MAmmoTH-VL":[91,104],"single-image":[92],"subset,":[93],"compare":[95],"models":[96,138],"curated":[100],"data,":[101,105],"external":[107],"open-weight":[108],"frontier":[109],"VLMs.":[110],"On":[111],"controlled":[113],"20-evaluation":[114],"set":[115],"14":[117],"VLMs":[118],"at":[119,194,199,212,234,240],"1B-4B":[120],"activated":[121],"parameters,":[122],"hold":[124],"fixed":[127],"with":[128,188],"per-model":[130],"regression,":[131],"separating":[132],"from":[134,228],"quality,":[135],"price":[137],"FLOPs":[140],"per":[141,166],"answer.":[143],"Curation":[144,174],"buys":[145,176,209],"35x":[147],"Cost-of-Pass":[148],"advantage":[149],"over":[150,182],"most":[152],"verbose":[153,252],"4B":[154],"comparator":[155],"(Qwen3.5-4B)":[156],"within":[157],"$\\sim$1":[158],"pp":[159,193,198],"of":[160,230,238],"accuracy":[161,180,211],"(0.41":[162],"vs":[163,170],"14.58":[164],"TFLOPs":[165],"answer;":[168],"0.691":[169],"0.704":[171],"mean":[172],"accuracy).":[173],"also":[175],"+17.55-percentage-point":[178],"matched-length":[179],"gain":[181],"uncurated":[184],"baseline":[185],"grows":[187],"scale":[190],"(from":[191],"+16.7":[192],"1B":[195],"+21.2":[197],"4B).":[200],"This":[201],"improvement":[203],"concedes":[204],"no":[205,210],"quality:":[206],"generic":[207],"verbosity":[208,222],"any":[213],"capability":[214,232],"or":[215],"scale,":[216],"window":[219],"where":[220],"reasoning-structured":[221],"still":[223],"earns":[224],"its":[225],"tokens":[226],"shrinks":[227],"4":[229],"8":[231,239],"groups":[233],"2B":[235],"1":[237],"4B.":[241],"Per":[242],"example,":[243],"concise":[245],"even":[247],"reaches":[248],"answers":[250],"reasoning":[253,257],"misses,":[255],"marking":[256],"distinct":[260],"target":[262],"rather":[263],"than":[264],"something":[265],"gives":[267],"up.":[268],"this":[272],"regime":[273],"tokens-per-correct":[276],"problem,":[277],"lever":[282],"targets":[284],"directly.":[286]},"counts_by_year":[],"updated_date":"2026-07-28T07:46:37.118299","created_date":"2026-06-26T00:00:00"}