{"id":"https://openalex.org/W7204190163","doi":"https://doi.org/10.48550/arxiv.2608.21614","title":"SAEM: Stage-Aware Expert Management for Memory-Efficient MoE Inference in Chain-of-Thought Reasoning","display_name":"SAEM: Stage-Aware Expert Management for Memory-Efficient MoE Inference in Chain-of-Thought Reasoning","publication_year":2026,"publication_date":"2026-08-21","ids":{"openalex":"https://openalex.org/W7204190163","doi":"https://doi.org/10.48550/arxiv.2608.21614"},"language":"en","primary_location":{"id":"pmh:oai:arXiv.org:2608.21614","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2608.21614","pdf_url":"https://arxiv.org/pdf/2608.21614","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/2608.21614","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5148146566","display_name":"Yujie Zhang","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":"Zhang, Yujie","raw_affiliation_strings":["School of Computing , National University of Singapore , Singapore"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"School of Computing , National University of Singapore , Singapore","institution_ids":["https://openalex.org/I165932596"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5148091702","display_name":"Bin Gao","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":"Gao, Bin","raw_affiliation_strings":["School of Computing , National University of Singapore , Singapore"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"School of Computing , National University of Singapore , Singapore","institution_ids":["https://openalex.org/I165932596"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5049237676","display_name":"Tulika Mitra","orcid":"https://orcid.org/0000-0003-4136-4188","observed_orcids":["https://orcid.org/0000-0003-4136-4188"]},"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":"Mitra, Tulika","raw_affiliation_strings":["School of Computing , National University of Singapore , Singapore"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"School of Computing , National University of Singapore , Singapore","institution_ids":["https://openalex.org/I165932596"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":1,"corresponding_author_ids":[],"corresponding_institution_ids":["https://openalex.org/I165932596"],"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/T10054","display_name":"Parallel Computing and Optimization Techniques","score":0.9355000257492065,"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/T10054","display_name":"Parallel Computing and Optimization Techniques","score":0.9355000257492065,"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":0.020400000736117363,"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":0.012400000356137753,"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/chain-of-thought-reasoning","display_name":"chain-of-thought reasoning","score":0.7820000052452087},{"id":"https://openalex.org/keywords/mixture-of-experts","display_name":"mixture of experts","score":0.871999979019165},{"id":"https://openalex.org/keywords/memory-efficient-inference","display_name":"memory-efficient inference","score":0.6150000095367432},{"id":"https://openalex.org/keywords/gpu-memory-constraints","display_name":"GPU memory constraints","score":0.5080000162124634}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.8348000049591064},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.7141000032424927},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.5198000073432922},{"id":"https://openalex.org/C82876162","wikidata":"https://www.wikidata.org/wiki/Q17096504","display_name":"Latency (audio)","level":2,"score":0.503600001335144},{"id":"https://openalex.org/C58328972","wikidata":"https://www.wikidata.org/wiki/Q184609","display_name":"Expert system","level":2,"score":0.48579999804496765},{"id":"https://openalex.org/C165696696","wikidata":"https://www.wikidata.org/wiki/Q11287","display_name":"Exploit","level":2,"score":0.48539999127388},{"id":"https://openalex.org/C26517878","wikidata":"https://www.wikidata.org/wiki/Q228039","display_name":"Key (lock)","level":2,"score":0.46209999918937683},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.4456999897956848},{"id":"https://openalex.org/C105002631","wikidata":"https://www.wikidata.org/wiki/Q4833645","display_name":"Subject-matter expert","level":3,"score":0.43160000443458557},{"id":"https://openalex.org/C86827895","wikidata":"https://www.wikidata.org/wiki/Q7098582","display_name":"Opportunistic reasoning","level":4,"score":0.428600013256073},{"id":"https://openalex.org/C74193536","wikidata":"https://www.wikidata.org/wiki/Q574844","display_name":"Kernel (algebra)","level":2,"score":0.4171000123023987},{"id":"https://openalex.org/C48145219","wikidata":"https://www.wikidata.org/wiki/Q1335365","display_name":"Security token","level":2,"score":0.39469999074935913},{"id":"https://openalex.org/C46743427","wikidata":"https://www.wikidata.org/wiki/Q1341685","display_name":"Inference engine","level":3,"score":0.3603000044822693},{"id":"https://openalex.org/C125411270","wikidata":"https://www.wikidata.org/wiki/Q18653","display_name":"Encoding (memory)","level":2,"score":0.35740000009536743},{"id":"https://openalex.org/C2781181686","wikidata":"https://www.wikidata.org/wiki/Q4226068","display_name":"Coherence (philosophical gambling strategy)","level":2,"score":0.32670000195503235},{"id":"https://openalex.org/C19768560","wikidata":"https://www.wikidata.org/wiki/Q320727","display_name":"Dependency (UML)","level":2,"score":0.3107999861240387},{"id":"https://openalex.org/C89288958","wikidata":"https://www.wikidata.org/wiki/Q7301504","display_name":"Reasoning system","level":2,"score":0.30970001220703125},{"id":"https://openalex.org/C129916263","wikidata":"https://www.wikidata.org/wiki/Q1141183","display_name":"Backward chaining","level":4,"score":0.304500013589859},{"id":"https://openalex.org/C124101348","wikidata":"https://www.wikidata.org/wiki/Q172491","display_name":"Data mining","level":1,"score":0.301800012588501},{"id":"https://openalex.org/C161301231","wikidata":"https://www.wikidata.org/wiki/Q3478658","display_name":"Knowledge representation and reasoning","level":2,"score":0.29919999837875366},{"id":"https://openalex.org/C16311509","wikidata":"https://www.wikidata.org/wiki/Q4148050","display_name":"Dependency graph","level":3,"score":0.2897999882698059},{"id":"https://openalex.org/C102600418","wikidata":"https://www.wikidata.org/wiki/Q6517507","display_name":"Legal expert system","level":3,"score":0.28679999709129333},{"id":"https://openalex.org/C37335422","wikidata":"https://www.wikidata.org/wiki/Q6888134","display_name":"Model-based reasoning","level":3,"score":0.2809999883174896},{"id":"https://openalex.org/C189950617","wikidata":"https://www.wikidata.org/wiki/Q937228","display_name":"Property (philosophy)","level":2,"score":0.27230000495910645},{"id":"https://openalex.org/C157170001","wikidata":"https://www.wikidata.org/wiki/Q4781507","display_name":"Applications of artificial intelligence","level":2,"score":0.26739999651908875},{"id":"https://openalex.org/C2777904410","wikidata":"https://www.wikidata.org/wiki/Q7397","display_name":"Software","level":2,"score":0.26600000262260437},{"id":"https://openalex.org/C115537543","wikidata":"https://www.wikidata.org/wiki/Q165596","display_name":"Cache","level":2,"score":0.26019999384880066},{"id":"https://openalex.org/C44291984","wikidata":"https://www.wikidata.org/wiki/Q1074173","display_name":"Question answering","level":2,"score":0.2599000036716461},{"id":"https://openalex.org/C83725634","wikidata":"https://www.wikidata.org/wiki/Q7268699","display_name":"Qualitative reasoning","level":2,"score":0.2538999915122986},{"id":"https://openalex.org/C98045186","wikidata":"https://www.wikidata.org/wiki/Q205663","display_name":"Process (computing)","level":2,"score":0.2508000135421753}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:oai:arXiv.org:2608.21614","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2608.21614","pdf_url":"https://arxiv.org/pdf/2608.21614","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.2608.21614","doi":"https://doi.org/10.48550/arxiv.2608.21614","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2608.21614","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":null,"license_id":null,"version":null,"is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"Preprint"}],"best_oa_location":{"id":"pmh:oai:arXiv.org:2608.21614","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2608.21614","pdf_url":"https://arxiv.org/pdf/2608.21614","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":[],"sustainable_development_goals_aurora":[],"study_designs":[],"awards":[{"id":"https://openalex.org/G3882347383","display_name":null,"funder_award_id":"Tier 3","funder_id":"https://openalex.org/F4320320751","funder_display_name":"Ministry of Education - Singapore"},{"id":"https://openalex.org/G4015358484","display_name":null,"funder_award_id":"CRP23-2019-0003","funder_id":"https://openalex.org/F4320320709","funder_display_name":"National Research Foundation Singapore"},{"id":"https://openalex.org/G4773786609","display_name":null,"funder_award_id":"NRF-CRP23-2019-0003","funder_id":"https://openalex.org/F4320320671","funder_display_name":"National Research Foundation"},{"id":"https://openalex.org/G6614349337","display_name":null,"funder_award_id":"CRP23-2019-0003","funder_id":"https://openalex.org/F4320320671","funder_display_name":"National Research Foundation"},{"id":"https://openalex.org/G8039525913","display_name":null,"funder_award_id":"NRF-CRP23-2019-0003","funder_id":"https://openalex.org/F4320320709","funder_display_name":"National Research Foundation Singapore"},{"id":"https://openalex.org/G913999031","display_name":null,"funder_award_id":"MOE-MOET32024-0003","funder_id":"https://openalex.org/F4320320751","funder_display_name":"Ministry of Education - Singapore"}],"funders":[{"id":"https://openalex.org/F4320320671","display_name":"National Research Foundation","ror":"https://ror.org/05s0g1g46"},{"id":"https://openalex.org/F4320320709","display_name":"National Research Foundation Singapore","ror":"https://ror.org/03cpyc314"},{"id":"https://openalex.org/F4320320751","display_name":"Ministry of Education - Singapore","ror":"https://ror.org/01kcva023"},{"id":"https://openalex.org/F4320322724","display_name":"Ministry of Education, India","ror":"https://ror.org/048xjjh50"}],"has_content":{"pdf":true,"grobid_xml":true},"content_urls":{"pdf":"https://content.openalex.org/works/W7204190163.pdf","grobid_xml":"https://content.openalex.org/works/W7204190163.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":["https://openalex.org/W7154695310","https://openalex.org/W7162513836","https://openalex.org/W4391272939","https://openalex.org/W7164891892","https://openalex.org/W7172439849","https://openalex.org/W4414727549","https://openalex.org/W7166851804","https://openalex.org/W7162524418","https://openalex.org/W7222956968","https://openalex.org/W4404308452"],"abstract_inverted_index":{"Chain-of-thought":[0],"(CoT)":[1],"prompting":[2],"improves":[3],"LLM":[4],"reasoning":[5,61,92,126],"by":[6],"decomposing":[7],"complex":[8],"problems":[9],"into":[10],"intermediate":[11],"steps,":[12],"but":[13],"its":[14],"sequential":[15],"nature":[16],"increases":[17],"decoding":[18],"latency":[19],"and":[20,41,65,78,95,111,119,124,140],"memory":[21,40],"usage.":[22],"Mixture-of-Experts":[23],"(MoE)":[24],"models":[25],"scale":[26],"capacity":[27],"through":[28],"sparse":[29],"expert":[30,35,67,102],"activation,":[31],"yet":[32],"their":[33],"full":[34],"weights":[36],"often":[37],"exceed":[38],"GPU":[39,145],"require":[42],"costly":[43],"GPU-CPU":[44],"transfers.":[45],"Existing":[46],"runtimes":[47],"treat":[48],"all":[49],"tokens":[50],"uniformly,":[51],"overlooking":[52],"a":[53,85],"key":[54],"structural":[55],"property":[56],"of":[57,159],"CoT":[58,165],"traces:":[59],"consecutive":[60],"stages":[62],"exhibit":[63],"coherent":[64],"predictable":[66],"activation":[68,98],"patterns.":[69],"Ignoring":[70],"this":[71],"stage-level":[72,97],"regularity":[73],"leads":[74],"to":[75,100,115,148],"inefficient":[76],"caching":[77,139],"unnecessary":[79],"data":[80,117,152],"movement.":[81],"We":[82],"propose":[83],"SAEM,":[84],"stage-aware":[86,106],"MoE":[87,162],"inference":[88,163],"runtime":[89],"that":[90],"detects":[91],"stage":[93],"boundaries":[94],"exploits":[96],"coherence":[99],"guide":[101],"placement.":[103],"SAEM":[104,128],"combines":[105],"caching,":[107],"expert-aligned":[108],"token":[109],"repacking,":[110],"in-situ":[112],"CPU":[113],"execution":[114],"reduce":[116],"transfer":[118],"kernel":[120],"fragmentation.":[121],"On":[122],"mathematical":[123],"scientific":[125],"workloads,":[127],"achieves":[129],"an":[130],"average":[131],"1.33x":[132],"throughput":[133],"improvement":[134],"over":[135],"the":[136,154,157],"strongest":[137],"state-of-the-art":[138],"offloading":[141],"baselines":[142],"under":[143],"constrained":[144],"memory,":[146],"rising":[147],"1.54x":[149],"when":[150],"calibration":[151],"matches":[153],"workload,":[155],"demonstrating":[156],"effectiveness":[158],"stage-aware,":[160],"locality-driven":[161],"for":[164],"reasoning.":[166]},"counts_by_year":[],"updated_date":"2026-10-28T10:33:24.094632","created_date":"2026-08-26T00:00:00"}