{"id":"https://openalex.org/W7166880924","doi":"https://doi.org/10.18653/v1/2026.acl-long.1496","title":"Self-SoftCoT: A Self-Consistent Framework via Position-Aware Latent Space Reinforcement Learning","display_name":"Self-SoftCoT: A Self-Consistent Framework via Position-Aware Latent Space Reinforcement Learning","publication_year":2026,"publication_date":"2026-01-01","ids":{"openalex":"https://openalex.org/W7166880924","doi":"https://doi.org/10.18653/v1/2026.acl-long.1496"},"language":null,"primary_location":{"id":"doi:10.18653/v1/2026.acl-long.1496","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2026.acl-long.1496","pdf_url":"https://aclanthology.org/2026.acl-long.1496.pdf","source":null,"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)","raw_type":"proceedings-article"},"type":"conference-paper","indexed_in":["crossref"],"open_access":{"is_oa":true,"oa_status":"gold","oa_url":"https://aclanthology.org/2026.acl-long.1496.pdf","any_repository_has_fulltext":null},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5139764718","display_name":"Liangliang Dong","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Liangliang Dong","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5010380116","display_name":"Lianlei Shan","orcid":"https://orcid.org/0000-0002-4648-8246"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Lianlei Shan","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5085846056","display_name":"Shuaimin Li","orcid":"https://orcid.org/0000-0002-8368-916X"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Shuaimin Li","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":true,"cited_by_count":0,"citation_normalized_percentile":{"value":0.8694276,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"32393","last_page":"32414"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T11307","display_name":"Domain Adaptation and Few-Shot Learning","score":0.20329999923706055,"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/T11307","display_name":"Domain Adaptation and Few-Shot Learning","score":0.20329999923706055,"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":0.1785999983549118,"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/T10036","display_name":"Advanced Neural Network Applications","score":0.07129999995231628,"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/reinforcement-learning","display_name":"Reinforcement learning","score":0.44040000438690186},{"id":"https://openalex.org/keywords/space","display_name":"Space (punctuation)","score":0.39719998836517334},{"id":"https://openalex.org/keywords/stability","display_name":"Stability (learning theory)","score":0.32499998807907104},{"id":"https://openalex.org/keywords/feature","display_name":"Feature (linguistics)","score":0.2973000109195709},{"id":"https://openalex.org/keywords/key","display_name":"Key (lock)","score":0.26579999923706055}],"concepts":[{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.5753999948501587},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.5702000260353088},{"id":"https://openalex.org/C97541855","wikidata":"https://www.wikidata.org/wiki/Q830687","display_name":"Reinforcement learning","level":2,"score":0.44040000438690186},{"id":"https://openalex.org/C2778572836","wikidata":"https://www.wikidata.org/wiki/Q380933","display_name":"Space (punctuation)","level":2,"score":0.39719998836517334},{"id":"https://openalex.org/C112972136","wikidata":"https://www.wikidata.org/wiki/Q7595718","display_name":"Stability (learning theory)","level":2,"score":0.32499998807907104},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.3222000002861023},{"id":"https://openalex.org/C2776401178","wikidata":"https://www.wikidata.org/wiki/Q12050496","display_name":"Feature (linguistics)","level":2,"score":0.2973000109195709},{"id":"https://openalex.org/C26517878","wikidata":"https://www.wikidata.org/wiki/Q228039","display_name":"Key (lock)","level":2,"score":0.26579999923706055},{"id":"https://openalex.org/C33923547","wikidata":"https://www.wikidata.org/wiki/Q395","display_name":"Mathematics","level":0,"score":0.26089999079704285},{"id":"https://openalex.org/C2780791683","wikidata":"https://www.wikidata.org/wiki/Q846785","display_name":"Action (physics)","level":2,"score":0.2526000142097473},{"id":"https://openalex.org/C99498987","wikidata":"https://www.wikidata.org/wiki/Q2210247","display_name":"Noise (video)","level":3,"score":0.2524000108242035}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.18653/v1/2026.acl-long.1496","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2026.acl-long.1496","pdf_url":"https://aclanthology.org/2026.acl-long.1496.pdf","source":null,"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)","raw_type":"proceedings-article"}],"best_oa_location":{"id":"doi:10.18653/v1/2026.acl-long.1496","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2026.acl-long.1496","pdf_url":"https://aclanthology.org/2026.acl-long.1496.pdf","source":null,"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)","raw_type":"proceedings-article"},"sustainable_development_goals":[{"id":"https://metadata.un.org/sdg/10","display_name":"Reduced inequalities","score":0.5463197827339172}],"awards":[],"funders":[{"id":"https://openalex.org/F4320314997","display_name":"Strong","ror":"https://ror.org/041vyzr56"},{"id":"https://openalex.org/F4320322300","display_name":"Jilin University","ror":"https://ror.org/00js3aw79"}],"has_content":{"pdf":true,"grobid_xml":true},"content_urls":{"pdf":"https://content.openalex.org/works/W7166880924.pdf","grobid_xml":"https://content.openalex.org/works/W7166880924.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"While":[0],"Chain-of-Thought":[1],"(CoT)":[2],"reasoning":[3,32,118,127],"empowers":[4],"Large":[5],"Language":[6],"Models":[7],"(LLMs)":[8],"to":[9,72,103,111,145,159,169],"tackle":[10],"complex":[11,52],"tasks,":[12],"its":[13],"reliance":[14],"on":[15,46,116],"discrete":[16],"token":[17],"decoding":[18],"imposes":[19],"an":[20],"inherent":[21],"Discreteness":[22],"Bottleneck,":[23],"limiting":[24],"expressiveness":[25],"within":[26],"a":[27,64,69,83],"restricted":[28],"vocabulary":[29],"space.Existing":[30],"continuous":[31],"approaches,":[33],"such":[34],"as":[35],"SoftCoT":[36],"(Xu":[37],"et":[38,135],"al.,":[39,136],"2025),":[40],"mitigate":[41,112],"this":[42],"but":[43],"typically":[44],"rely":[45],"external":[47,80],"auxiliary":[48],"models,":[49],"resulting":[50],"in":[51],"deployment":[53],"and":[54,75,106],"fractured":[55],"inference":[56],"pipelines.To":[57],"address":[58],"these":[59],"challenges,":[60],"we":[61,89,96],"propose":[62],"Self-SoftCoT,":[63],"self-contained":[65],"framework":[66],"that":[67,121],"enables":[68],"frozen":[70,130],"LLM":[71],"internally":[73],"generate":[74],"consume":[76],"latent":[77,91],"thoughts":[78],"without":[79],"assistants.By":[81],"establishing":[82],"singlestream":[84],"\"Thinking":[85],"\u2192":[86],"Speaking\"":[87],"closed-loop,":[88],"decouple":[90],"planning":[92],"from":[93,157,167],"explicit":[94],"generation.Furthermore,":[95],"adopt":[97],"Group":[98],"Sequence":[99],"Policy":[100],"Optimization":[101],"(GSPO)":[102],"stabilize":[104],"learning":[105],"employ":[107],"Position-Aware":[108],"Independent":[109],"Projection":[110],"representation":[113],"homogenization.Experimental":[114],"results":[115],"five":[117],"benchmarks":[119],"demonstrate":[120],"our":[122,132],"method":[123],"significantly":[124],"improves":[125],"the":[126,147,154],"performance":[128,165],"of":[129],"LLMs.Specifically,":[131],"Qwen2.5-basedmodel":[133],"(Yang":[134],"2024)":[137,164],"uses":[138],"only":[139],"N":[140],"=":[141,151],"2":[142],"soft":[143],"tokens":[144],"outperform":[146],"Soft-CoT":[148],"baseline":[149],"(N":[150],"4),":[152],"improving":[153],"average":[155],"accuracy":[156],"75.06%":[158],"78.42%.Similarly,":[160],"LLaMA-3.1":[161],"(Llama":[162],"Team,":[163],"increases":[166],"70.52%":[168],"74.55%":[170],"1":[171],".":[172]},"counts_by_year":[],"updated_date":"2026-07-29T14:22:42.915294","created_date":"2026-07-02T00:00:00"}