{"id":"https://openalex.org/W7130690025","doi":"https://doi.org/10.48550/arxiv.2602.16839","title":"Training Large Reasoning Models Efficiently via Progressive Thought Encoding","display_name":"Training Large Reasoning Models Efficiently via Progressive Thought Encoding","publication_year":2026,"publication_date":"2026-02-18","ids":{"openalex":"https://openalex.org/W7130690025","doi":"https://doi.org/10.48550/arxiv.2602.16839"},"language":null,"primary_location":{"id":"pmh:doi:10.48550/arxiv.2602.16839","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":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"Article"},"type":"article","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/A5126482818","display_name":"Zeliang Zhang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhang, Zeliang","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5126518129","display_name":"Xiaodong Liu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Liu, Xiaodong","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5126522506","display_name":"Hao Cheng","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Cheng, Hao","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5126480052","display_name":"Hao Sun","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Sun, Hao","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5126479884","display_name":"Chenliang Xu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Xu, Chenliang","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5126518790","display_name":"Jianfeng Gao","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Gao, Jianfeng","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":0.0,"has_fulltext":false,"cited_by_count":0,"citation_normalized_percentile":{"value":0.19837207,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"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/T10462","display_name":"Reinforcement Learning in Robotics","score":0.2953000068664551,"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/T10462","display_name":"Reinforcement Learning in Robotics","score":0.2953000068664551,"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":0.19779999554157257,"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/T12026","display_name":"Explainable Artificial Intelligence (XAI)","score":0.1264999955892563,"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/encoding","display_name":"Encoding (memory)","score":0.7696999907493591},{"id":"https://openalex.org/keywords/scalability","display_name":"Scalability","score":0.5404999852180481},{"id":"https://openalex.org/keywords/cache","display_name":"Cache","score":0.5254999995231628},{"id":"https://openalex.org/keywords/decoding-methods","display_name":"Decoding methods","score":0.49549999833106995},{"id":"https://openalex.org/keywords/face","display_name":"Face (sociological concept)","score":0.4300999939441681},{"id":"https://openalex.org/keywords/training","display_name":"Training (meteorology)","score":0.29649999737739563}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.8341000080108643},{"id":"https://openalex.org/C125411270","wikidata":"https://www.wikidata.org/wiki/Q18653","display_name":"Encoding (memory)","level":2,"score":0.7696999907493591},{"id":"https://openalex.org/C48044578","wikidata":"https://www.wikidata.org/wiki/Q727490","display_name":"Scalability","level":2,"score":0.5404999852180481},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.5303000211715698},{"id":"https://openalex.org/C115537543","wikidata":"https://www.wikidata.org/wiki/Q165596","display_name":"Cache","level":2,"score":0.5254999995231628},{"id":"https://openalex.org/C57273362","wikidata":"https://www.wikidata.org/wiki/Q576722","display_name":"Decoding methods","level":2,"score":0.49549999833106995},{"id":"https://openalex.org/C2779304628","wikidata":"https://www.wikidata.org/wiki/Q3503480","display_name":"Face (sociological concept)","level":2,"score":0.4300999939441681},{"id":"https://openalex.org/C80444323","wikidata":"https://www.wikidata.org/wiki/Q2878974","display_name":"Theoretical computer science","level":1,"score":0.3116999864578247},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.302700012922287},{"id":"https://openalex.org/C2777211547","wikidata":"https://www.wikidata.org/wiki/Q17141490","display_name":"Training (meteorology)","level":2,"score":0.29649999737739563},{"id":"https://openalex.org/C189783530","wikidata":"https://www.wikidata.org/wiki/Q352090","display_name":"CPU cache","level":3,"score":0.2937000095844269},{"id":"https://openalex.org/C51632099","wikidata":"https://www.wikidata.org/wiki/Q3985153","display_name":"Training set","level":2,"score":0.27309998869895935},{"id":"https://openalex.org/C97541855","wikidata":"https://www.wikidata.org/wiki/Q830687","display_name":"Reinforcement learning","level":2,"score":0.26420000195503235},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.2612000107765198},{"id":"https://openalex.org/C2777027219","wikidata":"https://www.wikidata.org/wiki/Q1284190","display_name":"Constant (computer programming)","level":2,"score":0.25870001316070557}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:doi:10.48550/arxiv.2602.16839","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":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"Article"},{"id":"doi:10.48550/arxiv.2602.16839","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2602.16839","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":"pmh:doi:10.48550/arxiv.2602.16839","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":"cc-by","license_id":"https://openalex.org/licenses/cc-by","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],"reasoning":[1,43,69,153],"models":[2],"(LRMs)":[3],"excel":[4],"on":[5,95,103,127,135],"complex":[6],"problems":[7],"but":[8,155],"face":[9],"a":[10,52],"critical":[11],"barrier":[12],"to":[13,59,79,131],"efficiency:":[14],"reinforcement":[15],"learning":[16],"(RL)":[17],"training":[18,159],"requires":[19],"long":[20],"rollouts":[21],"for":[22],"outcome-based":[23],"rewards,":[24],"where":[25],"autoregressive":[26],"decoding":[27],"dominates":[28],"time":[29],"and":[30,44,101,121,165],"memory":[31,86,91,169],"usage.":[32],"While":[33],"sliding-window":[34],"cache":[35,141],"strategies":[36],"can":[37],"bound":[38],"memory,":[39],"they":[40],"disrupt":[41],"long-context":[42],"degrade":[45],"performance.":[46],"We":[47],"introduce":[48],"Progressive":[49,147],"Thought":[50,148],"Encoding,":[51],"parameter-efficient":[53],"fine-tuning":[54,120,126],"method":[55,114],"that":[56,146],"enables":[57],"LRMs":[58,124,161],"reason":[60],"effectively":[61],"under":[62,137,167],"fixed-size":[63,71],"caches.":[64],"By":[65],"progressively":[66],"encoding":[67],"intermediate":[68],"into":[70],"vector":[72],"representations,":[73],"our":[74,113],"approach":[75],"eliminates":[76],"the":[77,138],"need":[78],"backpropagate":[80],"through":[81],"full-cache":[82],"rollouts,":[83],"thereby":[84],"reducing":[85],"usage,":[87],"while":[88],"maintaining":[89],"constant":[90],"during":[92],"inference.":[93],"Experiments":[94],"three":[96],"models,":[97],"including":[98],"Qwen2.5-3B-Instruct,":[99],"Qwen2.5-7B-Instruct,":[100],"DeepSeek-R1-Distill-Llama-8B,":[102],"six":[104],"widely":[105],"used":[106],"challenging":[107],"mathematical":[108],"benchmarks":[109],"show":[110],"consistent":[111],"gains:":[112],"achieves":[115],"+19.3%":[116],"improvement":[117,134],"over":[118,123],"LoRA-based":[119],"+29.9%":[122],"without":[125],"average,":[128],"with":[129],"up":[130],"+23.4":[132],"accuracy":[133,154],"AIME2024/2025":[136],"same":[139],"tight":[140],"budgets.":[142],"These":[143],"results":[144],"demonstrate":[145],"Encoding":[149],"not":[150],"only":[151],"improves":[152],"also":[156],"makes":[157],"RL":[158],"of":[160],"substantially":[162],"more":[163],"efficient":[164],"scalable":[166],"real-world":[168],"constraints.":[170]},"counts_by_year":[],"updated_date":"2026-07-28T07:46:37.118299","created_date":"2026-02-21T00:00:00"}