{"id":"https://openalex.org/W4412888115","doi":"https://doi.org/10.18653/v1/2025.findings-acl.820","title":"Reward Generalization in RLHF: A Topological Perspective","display_name":"Reward Generalization in RLHF: A Topological Perspective","publication_year":2025,"publication_date":"2025-01-01","ids":{"openalex":"https://openalex.org/W4412888115","doi":"https://doi.org/10.18653/v1/2025.findings-acl.820"},"language":"en","primary_location":{"id":"doi:10.18653/v1/2025.findings-acl.820","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.findings-acl.820","pdf_url":"https://aclanthology.org/2025.findings-acl.820.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":"Findings of the Association for Computational Linguistics: ACL 2025","raw_type":"proceedings-article"},"type":"conference-paper","indexed_in":["crossref"],"open_access":{"is_oa":true,"oa_status":"gold","oa_url":"https://aclanthology.org/2025.findings-acl.820.pdf","any_repository_has_fulltext":null},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5017959135","display_name":"Tianyi Qiu","orcid":"https://orcid.org/0000-0002-5778-0592"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Tianyi Alex Qiu","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5087782595","display_name":"Fanzhi Zeng","orcid":"https://orcid.org/0000-0002-6771-9664"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Fanzhi Zeng","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5093946962","display_name":"Jiaming Ji","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Jiaming Ji","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5113839420","display_name":"Dong Yan","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Dong Yan","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5042816569","display_name":"Kaile Wang","orcid":"https://orcid.org/0000-0002-2964-4614"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Kaile Wang","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5100568107","display_name":"Jiayi Zhou","orcid":"https://orcid.org/0000-0001-8865-3070"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Jiayi Zhou","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5100314669","display_name":"Han Yang","orcid":"https://orcid.org/0009-0000-0414-9417"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Yang Han","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5101306100","display_name":"Josef Dai","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Josef Dai","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5109683693","display_name":"Xuehai Pan","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Xuehai Pan","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5090073634","display_name":"Yaodong Yang","orcid":"https://orcid.org/0000-0001-8132-5613"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Yaodong Yang","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":0,"citation_normalized_percentile":null,"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"15884","last_page":"15930"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T13644","display_name":"Control Systems in Engineering","score":0.1899999976158142,"subfield":{"id":"https://openalex.org/subfields/2210","display_name":"Mechanical Engineering"},"field":{"id":"https://openalex.org/fields/22","display_name":"Engineering"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},"topics":[{"id":"https://openalex.org/T13644","display_name":"Control Systems in Engineering","score":0.1899999976158142,"subfield":{"id":"https://openalex.org/subfields/2210","display_name":"Mechanical Engineering"},"field":{"id":"https://openalex.org/fields/22","display_name":"Engineering"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}}],"keywords":[{"id":"https://openalex.org/keywords/generalization","display_name":"Generalization","score":0.7508543729782104},{"id":"https://openalex.org/keywords/perspective","display_name":"Perspective (graphical)","score":0.6964571475982666},{"id":"https://openalex.org/keywords/topology","display_name":"Topology (electrical circuits)","score":0.529199481010437},{"id":"https://openalex.org/keywords/computer-science","display_name":"Computer science","score":0.4745232164859772},{"id":"https://openalex.org/keywords/mathematics","display_name":"Mathematics","score":0.347026526927948},{"id":"https://openalex.org/keywords/artificial-intelligence","display_name":"Artificial intelligence","score":0.2907564043998718},{"id":"https://openalex.org/keywords/combinatorics","display_name":"Combinatorics","score":0.1310890018939972},{"id":"https://openalex.org/keywords/mathematical-analysis","display_name":"Mathematical analysis","score":0.08004584908485413}],"concepts":[{"id":"https://openalex.org/C177148314","wikidata":"https://www.wikidata.org/wiki/Q170084","display_name":"Generalization","level":2,"score":0.7508543729782104},{"id":"https://openalex.org/C12713177","wikidata":"https://www.wikidata.org/wiki/Q1900281","display_name":"Perspective (graphical)","level":2,"score":0.6964571475982666},{"id":"https://openalex.org/C184720557","wikidata":"https://www.wikidata.org/wiki/Q7825049","display_name":"Topology (electrical circuits)","level":2,"score":0.529199481010437},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.4745232164859772},{"id":"https://openalex.org/C33923547","wikidata":"https://www.wikidata.org/wiki/Q395","display_name":"Mathematics","level":0,"score":0.347026526927948},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.2907564043998718},{"id":"https://openalex.org/C114614502","wikidata":"https://www.wikidata.org/wiki/Q76592","display_name":"Combinatorics","level":1,"score":0.1310890018939972},{"id":"https://openalex.org/C134306372","wikidata":"https://www.wikidata.org/wiki/Q7754","display_name":"Mathematical analysis","level":1,"score":0.08004584908485413}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.18653/v1/2025.findings-acl.820","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.findings-acl.820","pdf_url":"https://aclanthology.org/2025.findings-acl.820.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":"Findings of the Association for Computational Linguistics: ACL 2025","raw_type":"proceedings-article"}],"best_oa_location":{"id":"doi:10.18653/v1/2025.findings-acl.820","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.findings-acl.820","pdf_url":"https://aclanthology.org/2025.findings-acl.820.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":"Findings of the Association for Computational Linguistics: ACL 2025","raw_type":"proceedings-article"},"sustainable_development_goals":[],"awards":[{"id":"https://openalex.org/G2899022648","display_name":null,"funder_award_id":"62376013","funder_id":"https://openalex.org/F4320321001","funder_display_name":"National Natural Science Foundation of China"}],"funders":[{"id":"https://openalex.org/F4320321001","display_name":"National Natural Science Foundation of China","ror":"https://ror.org/01h0zpd94"}],"has_content":{"grobid_xml":true,"pdf":true},"content_urls":{"pdf":"https://content.openalex.org/works/W4412888115.pdf","grobid_xml":"https://content.openalex.org/works/W4412888115.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":["https://openalex.org/W2899084033","https://openalex.org/W2748952813","https://openalex.org/W4391375266","https://openalex.org/W1979597421","https://openalex.org/W2007980826","https://openalex.org/W2061531152","https://openalex.org/W3002753104","https://openalex.org/W2077600819","https://openalex.org/W2142036596","https://openalex.org/W2072657027"],"abstract_inverted_index":{"Existing":[0],"alignment":[1],"methods":[2],"share":[3],"a":[4,53,57],"common":[5],"topology":[6,29,72,188],"of":[7,45,59,73,102,123,177,194],"information":[8,12,74,89],"flow,":[9],"where":[10,159],"reward":[11,60,127,135,145,183],"is":[13,141,161],"collected":[14],"from":[15,65,137],"humans,":[16],"modeled":[17],"with":[18],"preference":[19,107,139],"learning,":[20],"and":[21,49,79,108],"used":[22],"to":[23,119,143,149,157],"tune":[24],"language":[25],"models.However,":[26],"this":[27],"shared":[28],"has":[30],"not":[31],"been":[32,39],"systematically":[33],"characterized,":[34],"nor":[35],"have":[36],"its":[37],"alternatives":[38],"thoroughly":[40],"explored,":[41],"leaving":[42],"the":[43,71,82,87,99,111,121,162,192],"problems":[44],"low":[46],"data":[47,195],"efficiency":[48],"unreliable":[50],"generalization":[51,61,184],"unaddressed.As":[52],"solution,":[54],"we":[55,85,114,133],"introduce":[56],"theory":[58],"in":[62],"reinforcement":[63],"learning":[64],"human":[66,106],"feedback":[67],"(RLHF),":[68],"focusing":[69],"on":[70,126,130,165],"flow":[75,90],"at":[76],"both":[77,131],"macro":[78,83],"micro":[80,112],"levels.At":[81],"level,":[84,113],"portray":[86],"RLHF":[88,100],"as":[91],"an":[92,173],"autoencoding":[93],"process":[94],"over":[95],"behavior":[96],"distributions,":[97],"formalizing":[98],"objective":[101],"distributional":[103],"consistency":[104],"between":[105],"model":[109,120],"behavior.At":[110],"present":[115],"induced":[116],"Bayesian":[117],"networks":[118],"impact":[122],"dataset":[124,163],"topologies":[125],"generalization.Combining":[128],"analysis":[129],"levels,":[132],"propose":[134],"modeling":[136],"tree-structured":[138],"information.It":[140],"shown":[142],"reduce":[144],"uncertainty":[146],"by":[147],"up":[148],"(log":[150],"n/":[151],"log":[152,153],"n)":[154],"times":[155],"compared":[156],"baselines,":[158,180],"n":[160],"size.Validation":[164],"three":[166],"NLP":[167],"tasks":[168],"shows":[169],"that":[170],"it":[171],"achieves":[172],"average":[174],"win":[175],"rate":[176],"65%":[178],"against":[179],"thus":[181],"improving":[182],"for":[185],"free":[186],"via":[187],"design,":[189],"while":[190],"reducing":[191],"amount":[193],"requiring":[196],"annotation.":[197]},"counts_by_year":[],"updated_date":"2026-07-14T23:27:15.235271","created_date":"2025-10-10T00:00:00"}