{"id":"https://openalex.org/W7135090750","doi":"https://doi.org/10.48550/arxiv.2603.10279","title":"Robust Post-Training for Generative Recommenders: Why Exponential Reward-Weighted SFT Outperforms RLHF","display_name":"Robust Post-Training for Generative Recommenders: Why Exponential Reward-Weighted SFT Outperforms RLHF","publication_year":2026,"publication_date":"2026-03-10","ids":{"openalex":"https://openalex.org/W7135090750","doi":"https://doi.org/10.48550/arxiv.2603.10279"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2603.10279","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2603.10279","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":null,"issn":null,"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":null,"license_id":null,"version":null,"is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"article"},"type":"preprint","indexed_in":["datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://doi.org/10.48550/arxiv.2603.10279","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5068682668","display_name":"Keertana V. Chidambaram","orcid":null},"institutions":[],"countries":[],"is_corresponding":true,"raw_author_name":"Chidambaram, Keertana","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5109431491","display_name":"Sanath Kumar Krishnamurthy","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Krishnamurthy, Sanath Kumar","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5111181554","display_name":"Qiuling Xu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Xu, Qiuling","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5128889450","display_name":"Ko-Jen Hsiao","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Hsiao, Ko-Jen","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5047314505","display_name":"Moumita Bhattacharya","orcid":"https://orcid.org/0000-0002-7836-4504"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Bhattacharya, Moumita","raw_affiliation_strings":[],"affiliations":[]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":5,"corresponding_author_ids":["https://openalex.org/A5068682668"],"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/T10203","display_name":"Recommender Systems and Techniques","score":0.9656000137329102,"subfield":{"id":"https://openalex.org/subfields/1710","display_name":"Information Systems"},"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/T10203","display_name":"Recommender Systems and Techniques","score":0.9656000137329102,"subfield":{"id":"https://openalex.org/subfields/1710","display_name":"Information Systems"},"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":0.008100000210106373,"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/T12101","display_name":"Advanced Bandit Algorithms Research","score":0.0032999999821186066,"subfield":{"id":"https://openalex.org/subfields/1803","display_name":"Management Science and Operations Research"},"field":{"id":"https://openalex.org/fields/18","display_name":"Decision Sciences"},"domain":{"id":"https://openalex.org/domains/2","display_name":"Social Sciences"}}],"keywords":[{"id":"https://openalex.org/keywords/hyperparameter","display_name":"Hyperparameter","score":0.5909000039100647},{"id":"https://openalex.org/keywords/recommender-system","display_name":"Recommender system","score":0.5684999823570251},{"id":"https://openalex.org/keywords/generative-grammar","display_name":"Generative grammar","score":0.5259000062942505},{"id":"https://openalex.org/keywords/exponential-function","display_name":"Exponential function","score":0.47279998660087585},{"id":"https://openalex.org/keywords/weighting","display_name":"Weighting","score":0.47189998626708984},{"id":"https://openalex.org/keywords/generative-model","display_name":"Generative model","score":0.423799991607666},{"id":"https://openalex.org/keywords/regularization","display_name":"Regularization (linguistics)","score":0.4212000072002411}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6972000002861023},{"id":"https://openalex.org/C8642999","wikidata":"https://www.wikidata.org/wiki/Q4171168","display_name":"Hyperparameter","level":2,"score":0.5909000039100647},{"id":"https://openalex.org/C557471498","wikidata":"https://www.wikidata.org/wiki/Q554950","display_name":"Recommender system","level":2,"score":0.5684999823570251},{"id":"https://openalex.org/C39890363","wikidata":"https://www.wikidata.org/wiki/Q36108","display_name":"Generative grammar","level":2,"score":0.5259000062942505},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.48980000615119934},{"id":"https://openalex.org/C151376022","wikidata":"https://www.wikidata.org/wiki/Q168698","display_name":"Exponential function","level":2,"score":0.47279998660087585},{"id":"https://openalex.org/C183115368","wikidata":"https://www.wikidata.org/wiki/Q856577","display_name":"Weighting","level":2,"score":0.47189998626708984},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.4702000021934509},{"id":"https://openalex.org/C167966045","wikidata":"https://www.wikidata.org/wiki/Q5532625","display_name":"Generative model","level":3,"score":0.423799991607666},{"id":"https://openalex.org/C2776135515","wikidata":"https://www.wikidata.org/wiki/Q17143721","display_name":"Regularization (linguistics)","level":2,"score":0.4212000072002411},{"id":"https://openalex.org/C55974624","wikidata":"https://www.wikidata.org/wiki/Q1188504","display_name":"Exponential family","level":2,"score":0.38019999861717224},{"id":"https://openalex.org/C75235859","wikidata":"https://www.wikidata.org/wiki/Q582659","display_name":"Exponential growth","level":2,"score":0.3573000133037567},{"id":"https://openalex.org/C2778775528","wikidata":"https://www.wikidata.org/wiki/Q5135432","display_name":"Closing (real estate)","level":2,"score":0.3499999940395355},{"id":"https://openalex.org/C26517878","wikidata":"https://www.wikidata.org/wiki/Q228039","display_name":"Key (lock)","level":2,"score":0.31209999322891235},{"id":"https://openalex.org/C120936955","wikidata":"https://www.wikidata.org/wiki/Q2155640","display_name":"Empirical research","level":2,"score":0.3093000054359436},{"id":"https://openalex.org/C63479239","wikidata":"https://www.wikidata.org/wiki/Q7353546","display_name":"Robustness (evolution)","level":3,"score":0.30230000615119934},{"id":"https://openalex.org/C2986087404","wikidata":"https://www.wikidata.org/wiki/Q15946010","display_name":"Online learning","level":2,"score":0.29109999537467957},{"id":"https://openalex.org/C124101348","wikidata":"https://www.wikidata.org/wiki/Q172491","display_name":"Data mining","level":1,"score":0.28949999809265137},{"id":"https://openalex.org/C21569690","wikidata":"https://www.wikidata.org/wiki/Q94702","display_name":"Collaborative filtering","level":3,"score":0.2635999917984009},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.25290000438690186}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2603.10279","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2603.10279","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":null,"issn":null,"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":null,"license_id":null,"version":null,"is_accepted":false,"is_published":null,"raw_source_name":null,"raw_type":"article"}],"best_oa_location":{"id":"doi:10.48550/arxiv.2603.10279","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2603.10279","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":null,"issn":null,"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":null,"license_id":null,"version":null,"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":{"Aligning":[0],"generative":[1],"recommender":[2],"systems":[3],"to":[4,35,70,99],"user":[5,37],"preferences":[6],"via":[7],"post-training":[8,23],"is":[9,55,97,107,181],"critical":[10],"for":[11,27,117,137],"closing":[12],"the":[13,75,95,112,125,151],"gap":[14,126],"between":[15],"next-item":[16],"prediction":[17],"and":[18,39,52,73,77,106,133,148,169,184],"actual":[19],"recommendation":[20],"quality.":[21],"Existing":[22],"methods":[24,31],"are":[25,50],"ill-suited":[26],"production-scale":[28],"systems:":[29],"RLHF":[30],"reward":[32,41,93,100,179],"hack":[33],"due":[34],"noisy":[36,121],"feedback":[38],"unreliable":[40],"models,":[42],"offline":[43],"RL":[44],"alternatives":[45],"require":[46],"propensity":[47,104],"scores":[48],"that":[49,80,124,144,177],"unavailable,":[51],"online":[53],"interaction":[54],"infeasible.":[56],"We":[57,110],"identify":[58],"exponential":[59,178],"reward-weighted":[60],"SFT":[61],"with":[62,130,156,162],"weights":[63],"$w":[64],"=":[65],"\\exp(r/\u03bb)$":[66],"as":[67],"uniquely":[68],"suited":[69],"this":[71,118],"setting,":[72],"provide":[74],"theoretical":[76,163],"empirical":[78],"foundations":[79],"explain":[81],"why.":[82],"By":[83],"optimizing":[84],"directly":[85],"on":[86,166],"observed":[87],"rewards":[88],"without":[89],"querying":[90],"a":[91,157],"learned":[92],"model,":[94],"method":[96],"immune":[98],"hacking,":[101],"requires":[102],"no":[103],"scores,":[105],"fully":[108],"offline.":[109],"prove":[111],"first":[113],"policy":[114],"improvement":[115],"guarantees":[116],"setting":[119],"under":[120],"rewards,":[122],"showing":[123],"scales":[127],"only":[128],"logarithmically":[129],"catalog":[131],"size":[132],"remains":[134],"informative":[135],"even":[136],"large":[138],"item":[139],"catalogs.":[140],"Crucially,":[141],"we":[142],"show":[143],"temperature":[145],"$\u03bb$":[146],"explicitly":[147],"quantifiably":[149],"controls":[150],"robustness-improvement":[152],"tradeoff,":[153],"providing":[154],"practitioners":[155],"single":[158],"interpretable":[159],"regularization":[160],"hyperparameter":[161],"grounding.":[164],"Experiments":[165],"three":[167],"open-source":[168],"one":[170],"proprietary":[171],"dataset":[172],"against":[173],"four":[174],"baselines":[175],"confirm":[176],"weighting":[180],"simple,":[182],"scalable,":[183],"consistently":[185],"outperforms":[186],"RLHF-based":[187],"alternatives.":[188]},"counts_by_year":[],"updated_date":"2026-03-13T14:25:03.468858","created_date":"2026-03-13T00:00:00"}