{"id":"https://openalex.org/W7154736846","doi":"https://doi.org/10.48550/arxiv.2604.14922","title":"LongAct: Harnessing Intrinsic Activation Patterns for Long-Context Reinforcement Learning","display_name":"LongAct: Harnessing Intrinsic Activation Patterns for Long-Context Reinforcement Learning","publication_year":2026,"publication_date":"2026-04-16","ids":{"openalex":"https://openalex.org/W7154736846","doi":"https://doi.org/10.48550/arxiv.2604.14922"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2604.14922","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2604.14922","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":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.2604.14922","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5133859588","display_name":"Bowen Ping","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Ping, Bowen","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5133881927","display_name":"Zijun Chen","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Chen, Zijun","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5065626573","display_name":"Tingfeng Hui","orcid":"https://orcid.org/0009-0006-3524-4904"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Hui, Tingfeng","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5123323373","display_name":"Qize Yu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Yu, Qize","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5102014574","display_name":"Chenxuan Li","orcid":"https://orcid.org/0000-0001-5421-9225"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Li, Chenxuan","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5133854179","display_name":"Junchi Yan","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Yan, Junchi","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5133849902","display_name":"Baobao Chang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Chang, Baobao","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":0.2345999926328659,"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":0.2345999926328659,"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":0.093299999833107,"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":0.0754999965429306,"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/reinforcement-learning","display_name":"Reinforcement learning","score":0.696399986743927},{"id":"https://openalex.org/keywords/boosting","display_name":"Boosting (machine learning)","score":0.5748000144958496},{"id":"https://openalex.org/keywords/salient","display_name":"Salient","score":0.5623000264167786},{"id":"https://openalex.org/keywords/exploit","display_name":"Exploit","score":0.5530999898910522},{"id":"https://openalex.org/keywords/key","display_name":"Key (lock)","score":0.4862000048160553},{"id":"https://openalex.org/keywords/generalization","display_name":"Generalization","score":0.4587000012397766},{"id":"https://openalex.org/keywords/representation","display_name":"Representation (politics)","score":0.39890000224113464}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6988999843597412},{"id":"https://openalex.org/C97541855","wikidata":"https://www.wikidata.org/wiki/Q830687","display_name":"Reinforcement learning","level":2,"score":0.696399986743927},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.6320000290870667},{"id":"https://openalex.org/C46686674","wikidata":"https://www.wikidata.org/wiki/Q466303","display_name":"Boosting (machine learning)","level":2,"score":0.5748000144958496},{"id":"https://openalex.org/C2780719617","wikidata":"https://www.wikidata.org/wiki/Q1030752","display_name":"Salient","level":2,"score":0.5623000264167786},{"id":"https://openalex.org/C165696696","wikidata":"https://www.wikidata.org/wiki/Q11287","display_name":"Exploit","level":2,"score":0.5530999898910522},{"id":"https://openalex.org/C26517878","wikidata":"https://www.wikidata.org/wiki/Q228039","display_name":"Key (lock)","level":2,"score":0.4862000048160553},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.47119998931884766},{"id":"https://openalex.org/C177148314","wikidata":"https://www.wikidata.org/wiki/Q170084","display_name":"Generalization","level":2,"score":0.4587000012397766},{"id":"https://openalex.org/C2776359362","wikidata":"https://www.wikidata.org/wiki/Q2145286","display_name":"Representation (politics)","level":3,"score":0.39890000224113464},{"id":"https://openalex.org/C196340769","wikidata":"https://www.wikidata.org/wiki/Q7698910","display_name":"Temporal difference learning","level":3,"score":0.3788999915122986},{"id":"https://openalex.org/C59404180","wikidata":"https://www.wikidata.org/wiki/Q17013334","display_name":"Feature learning","level":2,"score":0.34369999170303345},{"id":"https://openalex.org/C28855332","wikidata":"https://www.wikidata.org/wiki/Q198099","display_name":"Quantization (signal processing)","level":2,"score":0.34360000491142273},{"id":"https://openalex.org/C95623464","wikidata":"https://www.wikidata.org/wiki/Q1096149","display_name":"Classifier (UML)","level":2,"score":0.34119999408721924},{"id":"https://openalex.org/C199833920","wikidata":"https://www.wikidata.org/wiki/Q612536","display_name":"Vector quantization","level":2,"score":0.3353999853134155},{"id":"https://openalex.org/C125411270","wikidata":"https://www.wikidata.org/wiki/Q18653","display_name":"Encoding (memory)","level":2,"score":0.2694999873638153}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2604.14922","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2604.14922","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":"doi:10.48550/arxiv.2604.14922","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2604.14922","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":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":{"Reinforcement":[0],"Learning":[1],"(RL)":[2],"has":[3],"emerged":[4],"as":[5,96,162],"a":[6,87,111],"critical":[7],"driver":[8],"for":[9,100],"enhancing":[10],"the":[11,33,40,49,55,72,80,97,125,145],"reasoning":[12,84],"capabilities":[13],"of":[14,51,74],"Large":[15],"Language":[16],"Models":[17],"(LLMs).":[18],"While":[19],"recent":[20],"advancements":[21],"have":[22],"focused":[23],"on":[24,105,138,144,172],"reward":[25],"engineering":[26],"or":[27],"data":[28],"synthesis,":[29],"few":[30],"studies":[31,168],"exploit":[32],"model's":[34],"intrinsic":[35],"representation":[36],"characteristics":[37],"to":[38,117,178],"guide":[39],"training":[41],"process.":[42],"In":[43],"this":[44,106],"paper,":[45],"we":[46,90,108],"first":[47],"observe":[48],"presence":[50],"high-magnitude":[52,76],"activations":[53,77],"within":[54],"query":[56],"and":[57,79,141,164],"key":[58,177],"vectors":[59],"when":[60],"processing":[61],"long":[62],"contexts.":[63],"Drawing":[64],"inspiration":[65],"from":[66,115],"model":[67,102],"quantization":[68],"--":[69,78],"which":[70],"establishes":[71],"criticality":[73],"such":[75,161],"insight":[81],"that":[82,92,113,170],"long-context":[83,180],"inherently":[85],"exhibits":[86,151],"sparse":[88,119],"structure,":[89],"hypothesize":[91],"these":[93,129,173],"weights":[94,126],"serve":[95],"pivotal":[98],"drivers":[99],"effective":[101],"optimization.":[103],"Based":[104],"insight,":[107],"propose":[109],"LongAct,":[110],"strategy":[112],"shifts":[114],"uniform":[116],"saliency-guided":[118],"updates.":[120],"By":[121],"selectively":[122],"updating":[123],"only":[124],"associated":[127],"with":[128],"significant":[130],"activations,":[131],"LongAct":[132],"achieves":[133],"an":[134],"approximate":[135],"8%":[136],"improvement":[137],"LongBench":[139],"v2":[140],"enhances":[142],"generalization":[143],"RULER":[146],"benchmark.":[147],"Furthermore,":[148],"our":[149],"method":[150],"remarkable":[152],"universality,":[153],"consistently":[154],"boosting":[155],"performance":[156],"across":[157],"diverse":[158],"RL":[159],"algorithms":[160],"GRPO":[163],"DAPO.":[165],"Extensive":[166],"ablation":[167],"suggest":[169],"focusing":[171],"salient":[174],"features":[175],"is":[176],"unlocking":[179],"potential.":[181]},"counts_by_year":[],"updated_date":"2026-07-28T07:46:37.118299","created_date":"2026-04-18T00:00:00"}