{"id":"https://openalex.org/W7128491674","doi":"https://doi.org/10.48550/arxiv.2602.07374","title":"TernaryLM: Memory-Efficient Language Modeling via Native 1.5-Bit Quantization with Adaptive Layer-wise Scaling","display_name":"TernaryLM: Memory-Efficient Language Modeling via Native 1.5-Bit Quantization with Adaptive Layer-wise Scaling","publication_year":2026,"publication_date":"2026-02-07","ids":{"openalex":"https://openalex.org/W7128491674","doi":"https://doi.org/10.48550/arxiv.2602.07374"},"language":null,"primary_location":{"id":"pmh:doi:10.48550/arxiv.2602.07374","is_oa":true,"landing_page_url":null,"pdf_url":null,"source":{"id":"https://openalex.org/S4406922384","display_name":"Open MIND","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"host_organization_lineage_names":[],"type":"repository"},"license":"publisher-specific-oa","license_id":"https://openalex.org/licenses/publisher-specific-oa","version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"Article"},"type":"preprint","indexed_in":["datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":null,"any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5093481083","display_name":"Nisharg Nargund","orcid":"https://orcid.org/0009-0007-2046-4864"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Nargund, Nisharg","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5067601336","display_name":"Priyesh Shukla","orcid":"https://orcid.org/0000-0002-9868-8944"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Shukla, Priyesh","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":2,"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/T10036","display_name":"Advanced Neural Network Applications","score":0.22370000183582306,"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/T10036","display_name":"Advanced Neural Network Applications","score":0.22370000183582306,"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":0.18649999797344208,"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":0.14249999821186066,"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/language-model","display_name":"Language model","score":0.6638000011444092},{"id":"https://openalex.org/keywords/perplexity","display_name":"Perplexity","score":0.6509000062942505},{"id":"https://openalex.org/keywords/quantization","display_name":"Quantization (signal processing)","score":0.6082000136375427},{"id":"https://openalex.org/keywords/scaling","display_name":"Scaling","score":0.48330000042915344},{"id":"https://openalex.org/keywords/inference","display_name":"Inference","score":0.48260000348091125},{"id":"https://openalex.org/keywords/transformer","display_name":"Transformer","score":0.44269999861717224},{"id":"https://openalex.org/keywords/artificial-neural-network","display_name":"Artificial neural network","score":0.3919999897480011},{"id":"https://openalex.org/keywords/limiting","display_name":"Limiting","score":0.3797000050544739}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7131999731063843},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.6638000011444092},{"id":"https://openalex.org/C100279451","wikidata":"https://www.wikidata.org/wiki/Q372193","display_name":"Perplexity","level":3,"score":0.6509000062942505},{"id":"https://openalex.org/C28855332","wikidata":"https://www.wikidata.org/wiki/Q198099","display_name":"Quantization (signal processing)","level":2,"score":0.6082000136375427},{"id":"https://openalex.org/C99844830","wikidata":"https://www.wikidata.org/wiki/Q102441924","display_name":"Scaling","level":2,"score":0.48330000042915344},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.48260000348091125},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.44760000705718994},{"id":"https://openalex.org/C66322947","wikidata":"https://www.wikidata.org/wiki/Q11658","display_name":"Transformer","level":3,"score":0.44269999861717224},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.40540000796318054},{"id":"https://openalex.org/C50644808","wikidata":"https://www.wikidata.org/wiki/Q192776","display_name":"Artificial neural network","level":2,"score":0.3919999897480011},{"id":"https://openalex.org/C188198153","wikidata":"https://www.wikidata.org/wiki/Q1613840","display_name":"Limiting","level":2,"score":0.3797000050544739},{"id":"https://openalex.org/C174348530","wikidata":"https://www.wikidata.org/wiki/Q188635","display_name":"Bridging (networking)","level":2,"score":0.3788999915122986},{"id":"https://openalex.org/C185429906","wikidata":"https://www.wikidata.org/wiki/Q1130160","display_name":"Estimator","level":2,"score":0.36559998989105225},{"id":"https://openalex.org/C80444323","wikidata":"https://www.wikidata.org/wiki/Q2878974","display_name":"Theoretical computer science","level":1,"score":0.34279999136924744},{"id":"https://openalex.org/C2779899686","wikidata":"https://www.wikidata.org/wiki/Q571227","display_name":"Implicature","level":3,"score":0.34200000762939453},{"id":"https://openalex.org/C2776224158","wikidata":"https://www.wikidata.org/wiki/Q187931","display_name":"Phrase","level":2,"score":0.310699999332428},{"id":"https://openalex.org/C138236772","wikidata":"https://www.wikidata.org/wiki/Q25098575","display_name":"Edge device","level":3,"score":0.3003999888896942},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.29179999232292175},{"id":"https://openalex.org/C2780922921","wikidata":"https://www.wikidata.org/wiki/Q255189","display_name":"Paraphrase","level":2,"score":0.28519999980926514},{"id":"https://openalex.org/C61423126","wikidata":"https://www.wikidata.org/wiki/Q187432","display_name":"Scripting language","level":2,"score":0.28299999237060547},{"id":"https://openalex.org/C203005215","wikidata":"https://www.wikidata.org/wiki/Q79798","display_name":"Machine translation","level":2,"score":0.2808000147342682},{"id":"https://openalex.org/C134537474","wikidata":"https://www.wikidata.org/wiki/Q17144832","display_name":"Naturalness","level":2,"score":0.2750000059604645},{"id":"https://openalex.org/C77660490","wikidata":"https://www.wikidata.org/wiki/Q244916","display_name":"Intermediate language","level":3,"score":0.27320000529289246},{"id":"https://openalex.org/C64452783","wikidata":"https://www.wikidata.org/wiki/Q1524945","display_name":"Ternary operation","level":2,"score":0.25519999861717224},{"id":"https://openalex.org/C2984842247","wikidata":"https://www.wikidata.org/wiki/Q197536","display_name":"Deep neural networks","level":3,"score":0.25110000371932983}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:doi:10.48550/arxiv.2602.07374","is_oa":true,"landing_page_url":null,"pdf_url":null,"source":{"id":"https://openalex.org/S4406922384","display_name":"Open MIND","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"host_organization_lineage_names":[],"type":"repository"},"license":"publisher-specific-oa","license_id":"https://openalex.org/licenses/publisher-specific-oa","version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"Article"},{"id":"doi:10.48550/arxiv.2602.07374","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2602.07374","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":null,"license_id":null,"version":null,"is_accepted":false,"is_published":null,"raw_source_name":null,"raw_type":"article"}],"best_oa_location":{"id":"pmh:doi:10.48550/arxiv.2602.07374","is_oa":true,"landing_page_url":null,"pdf_url":null,"source":{"id":"https://openalex.org/S4406922384","display_name":"Open MIND","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"host_organization_lineage_names":[],"type":"repository"},"license":"publisher-specific-oa","license_id":"https://openalex.org/licenses/publisher-specific-oa","version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"Article"},"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":{"Large":[0],"language":[1,45],"models":[2,193],"(LLMs)":[3],"achieve":[4,171],"remarkable":[5],"performance":[6],"but":[7],"demand":[8],"substantial":[9],"computational":[10],"resources,":[11],"limiting":[12],"deployment":[13],"on":[14,79,100,157],"edge":[15],"devices":[16],"and":[17,66,129,191],"resource-constrained":[18],"environments.":[19],"We":[20,160],"present":[21],"TernaryLM,":[22],"a":[23,82,140],"132M-parameter":[24],"transformer":[25,168],"trained":[26,192],"natively":[27],"with":[28,81,97,126],"ternary":[29,137],"quantization":[30,50,173],"{-1,":[31],"0,":[32],"+1}":[33],"(log2(3)":[34],"~":[35],"1.58-bit":[36],"effective":[37],"precision),":[38],"achieving":[39],"significant":[40],"memory":[41,112],"reduction":[42,113],"without":[43],"sacrificing":[44],"modeling":[46],"capability.":[47],"Unlike":[48],"post-training":[49],"approaches":[51],"that":[52,152,166],"quantize":[53],"pre-trained":[54],"full-precision":[55],"models,":[56],"TernaryLM":[57],"learns":[58],"quantization-aware":[59],"representations":[60],"from":[61],"scratch":[62],"using":[63,105],"straight-through":[64],"estimators":[65],"adaptive":[67],"per-layer":[68],"scaling":[69],"factors.":[70],"Our":[71,189],"experiments":[72],"demonstrate:":[73],"(1)":[74],"validation":[75],"perplexity":[76],"of":[77,86,123,143],"58.42":[78],"TinyStories":[80],"cross-seed":[83],"standard":[84],"deviation":[85],"+/-":[87],"0.17":[88],"PPL,":[89],"confirming":[90],"stable":[91],"optimization;":[92],"(2)":[93],"strong":[94],"downstream":[95],"transfer":[96],"82.47%":[98],"F1":[99],"MRPC,":[101],"surpassing":[102],"DistilBERT":[103],"despite":[104],"55x":[106],"less":[107],"pretraining":[108],"data;":[109],"(3)":[110],"2.4x":[111],"(498":[114],"MB":[115,118],"vs":[116],"1,197":[117],"for":[119,147,177,185],"an":[120,131,181],"FP32":[121,149],"model":[122],"identical":[124],"architecture)":[125],"latency":[127],"parity;":[128],"(4)":[130],"implicit":[132],"regularization":[133],"effect":[134],"whereby":[135],"the":[136,148],"constraint":[138],"yields":[139],"train/val":[141],"ratio":[142],"1.05x":[144],"versus":[145,175],"3.51x":[146],"baseline,":[150],"demonstrating":[151],"discrete":[153],"weights":[154],"prevent":[155],"overfitting":[156],"small":[158],"corpora.":[159],"provide":[161],"layer-wise":[162],"sparsity":[163,174],"analysis":[164],"revealing":[165],"middle":[167],"layers":[169],"(L5-L9)":[170],"60-62%":[172],"45-55%":[176],"boundary":[178],"layers,":[179],"establishing":[180],"actionable":[182],"design":[183],"principle":[184],"non-uniform":[186],"precision":[187],"allocation.":[188],"implementation":[190],"are":[194],"publicly":[195],"available":[196],"at":[197],"https://github.com/1nisharg/TernaryLM-Memory-Efficient-Language-Modeling.":[198]},"counts_by_year":[],"updated_date":"2026-06-11T09:08:48.828518","created_date":"2026-02-11T00:00:00"}