{"id":"https://openalex.org/W7161957511","doi":"https://doi.org/10.48550/arxiv.2605.20369","title":"DEL: Digit Entropy Loss for Numerical Learning of Large Language Models","display_name":"DEL: Digit Entropy Loss for Numerical Learning of Large Language Models","publication_year":2026,"publication_date":"2026-05-19","ids":{"openalex":"https://openalex.org/W7161957511","doi":"https://doi.org/10.48550/arxiv.2605.20369"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2605.20369","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.20369","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":null,"issn":null,"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.2605.20369","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5136628721","display_name":"Zhaohui Zheng","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zheng, Zhaohui","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5136639563","display_name":"Chenhang He","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"He, Chenhang","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5136654318","display_name":"Shihao Wang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wang, Shihao","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5136711429","display_name":"Yuxuan Li","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Li, Yuxuan","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5136615521","display_name":"Ming-Ming Cheng","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Cheng, Ming-Ming","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5136692757","display_name":"Lei Zhang","orcid":"https://orcid.org/0009-0004-8106-5603"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhang, Lei","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/T10028","display_name":"Topic Modeling","score":0.16380000114440918,"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":0.16380000114440918,"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/T11948","display_name":"Machine Learning in Materials Science","score":0.14880000054836273,"subfield":{"id":"https://openalex.org/subfields/2505","display_name":"Materials Chemistry"},"field":{"id":"https://openalex.org/fields/25","display_name":"Materials Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},{"id":"https://openalex.org/T13523","display_name":"Mathematics, Computing, and Information Processing","score":0.08150000125169754,"subfield":{"id":"https://openalex.org/subfields/1703","display_name":"Computational Theory and Mathematics"},"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/entropy","display_name":"Entropy (arrow of time)","score":0.5321999788284302},{"id":"https://openalex.org/keywords/cross-entropy","display_name":"Cross entropy","score":0.4004000127315521},{"id":"https://openalex.org/keywords/discretization","display_name":"Discretization","score":0.3978999853134155},{"id":"https://openalex.org/keywords/numerical-analysis","display_name":"Numerical analysis","score":0.3910999894142151},{"id":"https://openalex.org/keywords/decimal","display_name":"Decimal","score":0.3677999973297119},{"id":"https://openalex.org/keywords/binary-number","display_name":"Binary number","score":0.35510000586509705},{"id":"https://openalex.org/keywords/supervised-learning","display_name":"Supervised learning","score":0.3504999876022339},{"id":"https://openalex.org/keywords/source-code","display_name":"Source code","score":0.33799999952316284},{"id":"https://openalex.org/keywords/conditional-entropy","display_name":"Conditional entropy","score":0.3260999917984009}],"concepts":[{"id":"https://openalex.org/C106301342","wikidata":"https://www.wikidata.org/wiki/Q4117933","display_name":"Entropy (arrow of time)","level":2,"score":0.5321999788284302},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.5078999996185303},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.4478999972343445},{"id":"https://openalex.org/C167981619","wikidata":"https://www.wikidata.org/wiki/Q1685498","display_name":"Cross entropy","level":3,"score":0.4004000127315521},{"id":"https://openalex.org/C73000952","wikidata":"https://www.wikidata.org/wiki/Q17007827","display_name":"Discretization","level":2,"score":0.3978999853134155},{"id":"https://openalex.org/C33923547","wikidata":"https://www.wikidata.org/wiki/Q395","display_name":"Mathematics","level":0,"score":0.3921999931335449},{"id":"https://openalex.org/C48753275","wikidata":"https://www.wikidata.org/wiki/Q11216","display_name":"Numerical analysis","level":2,"score":0.3910999894142151},{"id":"https://openalex.org/C65045869","wikidata":"https://www.wikidata.org/wiki/Q81365","display_name":"Decimal","level":2,"score":0.3677999973297119},{"id":"https://openalex.org/C48372109","wikidata":"https://www.wikidata.org/wiki/Q3913","display_name":"Binary number","level":2,"score":0.35510000586509705},{"id":"https://openalex.org/C136389625","wikidata":"https://www.wikidata.org/wiki/Q334384","display_name":"Supervised learning","level":3,"score":0.3504999876022339},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.35019999742507935},{"id":"https://openalex.org/C43126263","wikidata":"https://www.wikidata.org/wiki/Q128751","display_name":"Source code","level":2,"score":0.33799999952316284},{"id":"https://openalex.org/C101721835","wikidata":"https://www.wikidata.org/wiki/Q813908","display_name":"Conditional entropy","level":3,"score":0.3260999917984009},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.3255000114440918},{"id":"https://openalex.org/C2776760102","wikidata":"https://www.wikidata.org/wiki/Q5139990","display_name":"Code (set theory)","level":3,"score":0.310699999332428},{"id":"https://openalex.org/C61797465","wikidata":"https://www.wikidata.org/wiki/Q1188986","display_name":"Term (time)","level":2,"score":0.2906999886035919},{"id":"https://openalex.org/C26517878","wikidata":"https://www.wikidata.org/wiki/Q228039","display_name":"Key (lock)","level":2,"score":0.2757999897003174},{"id":"https://openalex.org/C127349201","wikidata":"https://www.wikidata.org/wiki/Q753445","display_name":"Numerical integration","level":2,"score":0.27379998564720154},{"id":"https://openalex.org/C146849305","wikidata":"https://www.wikidata.org/wiki/Q370766","display_name":"Ground truth","level":2,"score":0.27160000801086426},{"id":"https://openalex.org/C94957134","wikidata":"https://www.wikidata.org/wiki/Q82990","display_name":"Numerical digit","level":2,"score":0.26930001378059387},{"id":"https://openalex.org/C500300565","wikidata":"https://www.wikidata.org/wiki/Q925667","display_name":"Computer simulation","level":2,"score":0.26429998874664307},{"id":"https://openalex.org/C8038995","wikidata":"https://www.wikidata.org/wiki/Q1152135","display_name":"Unsupervised learning","level":2,"score":0.26159998774528503},{"id":"https://openalex.org/C137836250","wikidata":"https://www.wikidata.org/wiki/Q984063","display_name":"Optimization problem","level":2,"score":0.2614000141620636},{"id":"https://openalex.org/C204160518","wikidata":"https://www.wikidata.org/wiki/Q122653","display_name":"Numeral system","level":2,"score":0.25429999828338623},{"id":"https://openalex.org/C9679016","wikidata":"https://www.wikidata.org/wiki/Q1417473","display_name":"Principle of maximum entropy","level":2,"score":0.25209999084472656},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.25040000677108765}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2605.20369","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.20369","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":null,"issn":null,"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.2605.20369","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.20369","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":null,"issn":null,"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":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Number":[0,38],"prediction":[1,212],"stands":[2],"as":[3],"a":[4,81,133,177],"fundamental":[5],"capability":[6],"of":[7,49,144],"large":[8],"language":[9],"models":[10],"(LLMs)":[11],"in":[12,116,209],"mathematical":[13,189],"problem-solving":[14],"and":[15,41,55,72,91,124,147,169,200,214],"code":[16],"generation.":[17],"The":[18],"widely":[19],"adopted":[20],"maximum":[21],"likelihood":[22],"estimation":[23],"(MLE)":[24],"for":[25,105],"LLM":[26,69],"training":[27],"is":[28],"not":[29],"tailored":[30],"to":[31,127,140,153,180],"number":[32,155,160,183],"prediction.":[33,161],"Recently,":[34],"penalty-driven":[35],"approaches,":[36],"e.g.,":[37],"Token":[39],"Loss":[40,103],"Discretized":[42],"Distance":[43],"Loss,":[44],"introduce":[45],"an":[46,65],"inductive":[47],"bias":[48],"numerical":[50,70,76,107,145,151,215],"distance":[51,93,138],"but":[52],"induce":[53],"over-sharpened":[54],"over-flattened":[56],"digit":[57,121,179],"distributions,":[58],"respectively.":[59],"In":[60],"this":[61],"paper,":[62],"we":[63,99],"make":[64],"in-depth":[66],"analysis":[67],"on":[68,187],"learning,":[71,108],"show":[73],"that":[74,203],"existing":[75],"learning":[77,152,174],"methods":[78],"conceptually":[79],"follow":[80],"criterion-distance":[82],"formulation,":[83],"where":[84],"the":[85,92,111,129,137,142,149,173,181],"criterion":[86],"term":[87,94,139],"represents":[88],"optimization":[89,115,131],"pattern":[90],"instills":[95],"geometric":[96],"prior.":[97],"Consequently,":[98],"present":[100],"Digit":[101],"Entropy":[102],"(DEL)":[104],"auto-regressive":[106],"which":[109],"reformulates":[110],"conventional":[112],"unsupervised":[113],"entropy":[114,130],"three":[117],"key":[118],"designs:":[119],"leveraging":[120],"conditional":[122],"probability":[123],"binary":[125],"cross-entropy":[126],"guide":[128],"into":[132],"supervised":[134],"manner;":[135],"deprecating":[136],"bypass":[141],"issue":[143],"distance;":[146],"generalizing":[148],"integer-based":[150],"floating-point":[154,182],"optimization,":[156],"enabling":[157],"more":[158],"accurate":[159],"Our":[162],"DEL":[163,204],"formulation":[164],"can":[165],"incorporate":[166],"integers,":[167],"decimals,":[168],"decimal":[170],"points,":[171],"expanding":[172],"objective":[175],"from":[176],"single":[178],"domain.":[184],"Experiments":[185],"conducted":[186],"seven":[188],"reasoning":[190],"benchmarks":[191],"with":[192],"four":[193],"representative":[194],"LLMs,":[195],"including":[196],"CodeLlama,":[197],"Mistral,":[198],"DeepSeek,":[199],"Qwen-2.5,":[201],"demonstrate":[202],"consistently":[205],"outperforms":[206],"its":[207],"counterparts":[208],"both":[210],"overall":[211],"accuracy":[213],"distance.":[216],"Source":[217],"codes":[218],"are":[219],"at":[220],"https://github.com/PolyU-VCLab/DEL":[221]},"counts_by_year":[],"updated_date":"2026-07-01T06:00:48.157686","created_date":"2026-05-22T00:00:00"}