{"id":"https://openalex.org/W7197032550","doi":"https://doi.org/10.48550/arxiv.2608.04771","title":"Fewer Tokens, Smaller Cache: Reward-Coordinated Efficient Reasoning","display_name":"Fewer Tokens, Smaller Cache: Reward-Coordinated Efficient Reasoning","publication_year":2026,"publication_date":"2026-08-05","ids":{"openalex":"https://openalex.org/W7197032550","doi":"https://doi.org/10.48550/arxiv.2608.04771"},"language":"en","primary_location":{"id":"pmh:oai:arXiv.org:2608.04771","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2608.04771","pdf_url":"https://arxiv.org/pdf/2608.04771","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":"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/2608.04771","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5140854880","display_name":"Qiyuan Zhu","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhu, Qiyuan","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5145140132","display_name":"Dezhi Li","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Li, Dezhi","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5128902934","display_name":"Pengyu Cheng","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Cheng, Pengyu","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5108908375","display_name":"Tianle Chen","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Chen, Tianle","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5127919888","display_name":"Jiacheng Wang","orcid":"https://orcid.org/0009-0002-0054-4684","observed_orcids":["https://orcid.org/0009-0002-0054-4684","https://orcid.org/0009-0004-3738-7509"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wang, Jiacheng","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5139849655","display_name":"Ruijie Shen","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Shen, Ruijie","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5121738655","display_name":"Hao Gu","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Gu, Hao","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5071566372","display_name":"Sida Lin","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Lin, Sida","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5138957861","display_name":"Zirui Liu","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Liu, Zirui","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5133846356","display_name":"Jiacheng Liu","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Liu, Jiacheng","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5021255269","display_name":"Sirui Han","orcid":"https://orcid.org/0000-0001-7303-0671","observed_orcids":["https://orcid.org/0000-0001-7303-0671"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Han, Sirui","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":true,"cited_by_count":2,"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":0.6880000233650208,"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":0.6880000233650208,"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":0.17759999632835388,"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/T11714","display_name":"Multimodal Machine Learning Applications","score":0.022099999710917473,"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"}}],"keywords":[{"id":"https://openalex.org/keywords/chain-of-thought-reasoning","display_name":"chain-of-thought reasoning","score":0.8610000014305115},{"id":"https://openalex.org/keywords/kv-cache-compression","display_name":"KV cache compression","score":0.9509999752044678},{"id":"https://openalex.org/keywords/early-stopping","display_name":"early stopping","score":0.7059999704360962},{"id":"https://openalex.org/keywords/inference-cost-reduction","display_name":"inference cost reduction","score":0.6240000128746033},{"id":"https://openalex.org/keywords/large-reasoning-models","display_name":"large reasoning models","score":0.7599999904632568}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7864999771118164},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.6480000019073486},{"id":"https://openalex.org/C98045186","wikidata":"https://www.wikidata.org/wiki/Q205663","display_name":"Process (computing)","level":2,"score":0.5730000138282776},{"id":"https://openalex.org/C115537543","wikidata":"https://www.wikidata.org/wiki/Q165596","display_name":"Cache","level":2,"score":0.5490000247955322},{"id":"https://openalex.org/C2779343474","wikidata":"https://www.wikidata.org/wiki/Q3109175","display_name":"Context (archaeology)","level":2,"score":0.5322999954223633},{"id":"https://openalex.org/C185429906","wikidata":"https://www.wikidata.org/wiki/Q1130160","display_name":"Estimator","level":2,"score":0.5271999835968018},{"id":"https://openalex.org/C180016635","wikidata":"https://www.wikidata.org/wiki/Q2712821","display_name":"Compression (physics)","level":2,"score":0.5058000087738037},{"id":"https://openalex.org/C82876162","wikidata":"https://www.wikidata.org/wiki/Q17096504","display_name":"Latency (audio)","level":2,"score":0.48820000886917114},{"id":"https://openalex.org/C78548338","wikidata":"https://www.wikidata.org/wiki/Q2493","display_name":"Data compression","level":2,"score":0.4691999852657318},{"id":"https://openalex.org/C28719098","wikidata":"https://www.wikidata.org/wiki/Q44946","display_name":"Point (geometry)","level":2,"score":0.462799996137619},{"id":"https://openalex.org/C65682993","wikidata":"https://www.wikidata.org/wiki/Q1056451","display_name":"Reflection (computer programming)","level":2,"score":0.4458000063896179},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.4027000069618225},{"id":"https://openalex.org/C81081738","wikidata":"https://www.wikidata.org/wiki/Q55542","display_name":"Lossless compression","level":3,"score":0.38119998574256897},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.3440000116825104},{"id":"https://openalex.org/C189783530","wikidata":"https://www.wikidata.org/wiki/Q352090","display_name":"CPU cache","level":3,"score":0.33559998869895935},{"id":"https://openalex.org/C48145219","wikidata":"https://www.wikidata.org/wiki/Q1335365","display_name":"Security token","level":2,"score":0.32499998807907104},{"id":"https://openalex.org/C25797200","wikidata":"https://www.wikidata.org/wiki/Q828137","display_name":"Compression ratio","level":3,"score":0.3188000023365021},{"id":"https://openalex.org/C13662910","wikidata":"https://www.wikidata.org/wiki/Q193139","display_name":"Trajectory","level":2,"score":0.31700000166893005},{"id":"https://openalex.org/C88871306","wikidata":"https://www.wikidata.org/wiki/Q7208287","display_name":"Point process","level":2,"score":0.2840999960899353},{"id":"https://openalex.org/C195344581","wikidata":"https://www.wikidata.org/wiki/Q2555318","display_name":"Automated reasoning","level":2,"score":0.27720001339912415},{"id":"https://openalex.org/C13481523","wikidata":"https://www.wikidata.org/wiki/Q412438","display_name":"Image compression","level":4,"score":0.2745000123977661},{"id":"https://openalex.org/C103057564","wikidata":"https://www.wikidata.org/wiki/Q4751139","display_name":"Analytic reasoning","level":3,"score":0.2687000036239624},{"id":"https://openalex.org/C80444323","wikidata":"https://www.wikidata.org/wiki/Q2878974","display_name":"Theoretical computer science","level":1,"score":0.2542000114917755}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:oai:arXiv.org:2608.04771","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2608.04771","pdf_url":"https://arxiv.org/pdf/2608.04771","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":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},{"id":"doi:10.48550/arxiv.2608.04771","doi":"https://doi.org/10.48550/arxiv.2608.04771","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2608.04771","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:2608.04771","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2608.04771","pdf_url":"https://arxiv.org/pdf/2608.04771","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":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},"sustainable_development_goals":[],"study_designs":[],"awards":[],"funders":[],"has_content":{"pdf":true,"grobid_xml":true},"content_urls":{"pdf":"https://content.openalex.org/works/W7197032550.pdf","grobid_xml":"https://content.openalex.org/works/W7197032550.grobid-xml"},"referenced_works_count":34,"referenced_works":["https://openalex.org/W3134642945","https://openalex.org/W3210904070","https://openalex.org/W4388926373","https://openalex.org/W4399414358","https://openalex.org/W4399912336","https://openalex.org/W4400141368","https://openalex.org/W4400611829","https://openalex.org/W4401202092","https://openalex.org/W4401306371","https://openalex.org/W4405766390","https://openalex.org/W4405902850","https://openalex.org/W4405957960","https://openalex.org/W4406892137","https://openalex.org/W4407208746","https://openalex.org/W4407231747","https://openalex.org/W4414281281","https://openalex.org/W4414633696","https://openalex.org/W4414939355","https://openalex.org/W4415063194","https://openalex.org/W4415330016","https://openalex.org/W4416035021","https://openalex.org/W4416365805","https://openalex.org/W4416915242","https://openalex.org/W4417102182","https://openalex.org/W4417300118","https://openalex.org/W6910423410","https://openalex.org/W6910586635","https://openalex.org/W6948369276","https://openalex.org/W6967523894","https://openalex.org/W7127094718","https://openalex.org/W7128484916","https://openalex.org/W7160928156","https://openalex.org/W7163178833","https://openalex.org/W7167587142"],"related_works":["https://openalex.org/W4415277444","https://openalex.org/W7138798841","https://openalex.org/W7166860003","https://openalex.org/W7114789567","https://openalex.org/W6929531631","https://openalex.org/W4416386252","https://openalex.org/W4417102263","https://openalex.org/W7211878501","https://openalex.org/W4415192111","https://openalex.org/W7126081047"],"abstract_inverted_index":{"Large":[0],"Reasoning":[1],"Models":[2],"(LRMs)":[3],"excel":[4],"on":[5,97,171],"complex":[6],"tasks":[7],"through":[8],"long":[9],"chain-of-thought":[10],"(CoT)":[11],"reasoning,":[12],"but":[13],"their":[14],"lengthy":[15],"intermediate":[16],"steps":[17,80,161],"cause":[18],"severe":[19],"overthinking":[20],"that":[21,153,174,183],"inflates":[22],"inference":[23],"cost.":[24],"KV-cache":[25,151],"compression":[26,44,93,152],"is":[27,94,188],"a":[28,36,60,102,123,132,137,168],"common":[29],"solution,":[30],"yet":[31],"existing":[32],"reasoning-oriented":[33],"methods":[34],"apply":[35],"uniform":[37],"policy":[38],"across":[39],"the":[40,51,56,69,87,98,106,114,155,186],"trajectory":[41],"and":[42,71,145,162,178,194,203],"judge":[43],"only":[45],"by":[46,201,206],"what":[47],"it":[48],"removes":[49],"from":[50],"cache.":[52],"Two":[53],"observations":[54],"point":[55],"other":[57],"way.":[58],"First,":[59],"reasoning":[61,187,192],"state's":[62],"tolerance":[63],"to":[64,108],"context":[65],"loss":[66],"varies":[67],"along":[68],"trajectory,":[70],"process":[72,125],"reward":[73],"tracks":[74],"it:":[75],"deleting":[76,86],"tokens":[77,173,200],"at":[78,90,159,164],"high-reward":[79,160],"preserves":[81],"accuracy":[82],"far":[83],"better":[84],"than":[85],"same":[88],"budget":[89],"random.":[91],"Second,":[92],"not":[95],"free":[96],"generation":[99],"side,":[100],"since":[101],"smaller":[103],"cache":[104,157],"leads":[105],"model":[107],"generate":[109],"more":[110],"tokens,":[111],"partly":[112],"canceling":[113],"saving.":[115],"Together":[116],"these":[117],"motivate":[118],"coordinating":[119],"both":[120],"sides":[121],"under":[122],"single":[124],"reward.":[126],"We":[127],"propose":[128],"ReCo":[129,197],"(Reward-Coordinated":[130],"Compression),":[131],"step-wise":[133],"framework":[134],"in":[135],"which":[136],"lightweight":[138],"process-reward":[139],"estimator":[140],"scores":[141],"each":[142],"completed":[143],"step":[144],"drives":[146],"three":[147,191],"components:":[148],"(1)":[149],"reward-adaptive":[150],"shrinks":[154],"retained":[156],"harder":[158],"less":[163],"low-reward":[165],"ones,":[166],"(2)":[167],"reward-banded":[169],"penalty":[170],"reflection":[172],"curbs":[175],"redundant":[176],"generation,":[177],"(3)":[179],"confidence-based":[180],"early":[181],"stopping":[182],"triggers":[184],"when":[185],"reliable.":[189],"Across":[190],"models":[193],"six":[195],"benchmarks,":[196],"reduces":[198],"generated":[199],"37%-65%":[202],"end-to-end":[204],"latency":[205],"2.08x-2.35x":[207],"over":[208],"Full":[209],"CoT,":[210],"all":[211],"while":[212],"largely":[213],"preserving":[214],"accuracy.":[215]},"counts_by_year":[{"year":2026,"cited_by_count":2}],"updated_date":"2026-11-07T11:25:07.637180","created_date":"2026-08-07T00:00:00"}