{"id":"https://openalex.org/W7223046115","doi":"https://doi.org/10.48550/arxiv.2610.24188","title":"On Reparameterizing the Score Function","display_name":"On Reparameterizing the Score Function","publication_year":2026,"publication_date":"2026-10-16","ids":{"openalex":"https://openalex.org/W7223046115","doi":"https://doi.org/10.48550/arxiv.2610.24188"},"language":"en","primary_location":{"id":"pmh:oai:arXiv.org:2610.24188","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2610.24188","pdf_url":"https://arxiv.org/pdf/2610.24188","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,"listed_in":[],"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":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},"volume":null,"type":"preprint","is_traditional":true,"indexed_in":["arxiv","datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://arxiv.org/pdf/2610.24188","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5086801110","display_name":"Mohammad Hossein Amani","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I5124864","display_name":"\u00c9cole Polytechnique F\u00e9d\u00e9rale de Lausanne","ror":"https://ror.org/02s376052","country_code":"CH","type":"education","lineage":["https://openalex.org/I2799323385","https://openalex.org/I5124864"]}],"countries":["CH"],"is_corresponding":false,"raw_author_name":"Amani, Mohammad Hossein","raw_affiliation_strings":["EPFL"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"EPFL","institution_ids":["https://openalex.org/I5124864"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5143844240","display_name":"Leonardo Bocchieri","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I5124864","display_name":"\u00c9cole Polytechnique F\u00e9d\u00e9rale de Lausanne","ror":"https://ror.org/02s376052","country_code":"CH","type":"education","lineage":["https://openalex.org/I2799323385","https://openalex.org/I5124864"]}],"countries":["CH"],"is_corresponding":false,"raw_author_name":"Bocchieri, Leonardo","raw_affiliation_strings":["EPFL"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"EPFL","institution_ids":["https://openalex.org/I5124864"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5078829915","display_name":"Maxime Peyrard","orcid":"https://orcid.org/0000-0003-4782-6603","observed_orcids":["https://orcid.org/0000-0003-4782-6603"]},"institutions":[{"id":"https://openalex.org/I1294671590","display_name":"Centre National de la Recherche Scientifique","ror":"https://ror.org/02feahw73","country_code":"FR","type":"government","lineage":["https://openalex.org/I1294671590"]},{"id":"https://openalex.org/I899635006","display_name":"Universit\u00e9 Grenoble Alpes","ror":"https://ror.org/02rx3b187","country_code":"FR","type":"education","lineage":["https://openalex.org/I899635006"]}],"countries":["FR"],"is_corresponding":false,"raw_author_name":"Peyrard, Maxime","raw_affiliation_strings":["CNRS , Univ. Grenoble Alpes"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"CNRS , Univ. Grenoble Alpes","institution_ids":["https://openalex.org/I1294671590","https://openalex.org/I899635006"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5063861960","display_name":"Cl\u00e9ment Hongler","orcid":"https://orcid.org/0000-0002-6224-2035","observed_orcids":["https://orcid.org/0000-0002-6224-2035"]},"institutions":[{"id":"https://openalex.org/I5124864","display_name":"\u00c9cole Polytechnique F\u00e9d\u00e9rale de Lausanne","ror":"https://ror.org/02s376052","country_code":"CH","type":"education","lineage":["https://openalex.org/I2799323385","https://openalex.org/I5124864"]}],"countries":["CH"],"is_corresponding":false,"raw_author_name":"Hongler, Cl\u00e9ment","raw_affiliation_strings":["EPFL","XENT Labs"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"EPFL","institution_ids":["https://openalex.org/I5124864"]},{"raw_affiliation_string":"XENT Labs","institution_ids":[]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5153518921","display_name":"Robert West","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I5124864","display_name":"\u00c9cole Polytechnique F\u00e9d\u00e9rale de Lausanne","ror":"https://ror.org/02s376052","country_code":"CH","type":"education","lineage":["https://openalex.org/I2799323385","https://openalex.org/I5124864"]}],"countries":["CH"],"is_corresponding":false,"raw_author_name":"West, Robert","raw_affiliation_strings":["EPFL"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"EPFL","institution_ids":["https://openalex.org/I5124864"]}]}],"institutions":[],"countries_distinct_count":2,"institutions_distinct_count":3,"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":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":0.9970999956130981,"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":0.9970999956130981,"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/T12794","display_name":"Adaptive Dynamic Programming Control","score":0.0006000000284984708,"subfield":{"id":"https://openalex.org/subfields/2207","display_name":"Control and Systems Engineering"},"field":{"id":"https://openalex.org/fields/22","display_name":"Engineering"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},{"id":"https://openalex.org/T11195","display_name":"Simulation Techniques and Applications","score":0.0003000000142492354,"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/score-function","display_name":"score function","score":0.5320000052452087},{"id":"https://openalex.org/keywords/credit-assignment","display_name":"credit assignment","score":0.7559999823570251},{"id":"https://openalex.org/keywords/variance-reduction","display_name":"variance reduction","score":0.828000009059906},{"id":"https://openalex.org/keywords/gradient-variance","display_name":"gradient variance","score":0.6520000100135803},{"id":"https://openalex.org/keywords/sparse-rewards","display_name":"sparse rewards","score":0.7009999752044678},{"id":"https://openalex.org/keywords/markov-decision-process","display_name":"Markov decision process","score":0.6959999799728394},{"id":"https://openalex.org/keywords/traveling-salesman-problem","display_name":"traveling salesman problem","score":0.7649999856948853},{"id":"https://openalex.org/keywords/autoregressive-transformer","display_name":"autoregressive transformers","score":0.6000000238418579}],"concepts":[{"id":"https://openalex.org/C33923547","wikidata":"https://www.wikidata.org/wiki/Q395","display_name":"Mathematics","level":0,"score":0.6966999769210815},{"id":"https://openalex.org/C159877910","wikidata":"https://www.wikidata.org/wiki/Q2202883","display_name":"Autoregressive model","level":2,"score":0.5684999823570251},{"id":"https://openalex.org/C185429906","wikidata":"https://www.wikidata.org/wiki/Q1130160","display_name":"Estimator","level":2,"score":0.5291000008583069},{"id":"https://openalex.org/C62644790","wikidata":"https://www.wikidata.org/wiki/Q3454689","display_name":"Variance reduction","level":3,"score":0.49619999527931213},{"id":"https://openalex.org/C2908647359","wikidata":"https://www.wikidata.org/wiki/Q2625603","display_name":"Population","level":2,"score":0.42579999566078186},{"id":"https://openalex.org/C149782125","wikidata":"https://www.wikidata.org/wiki/Q160039","display_name":"Econometrics","level":1,"score":0.39879998564720154},{"id":"https://openalex.org/C28826006","wikidata":"https://www.wikidata.org/wiki/Q33521","display_name":"Applied mathematics","level":1,"score":0.38580000400543213},{"id":"https://openalex.org/C98763669","wikidata":"https://www.wikidata.org/wiki/Q176645","display_name":"Markov chain","level":2,"score":0.382099986076355},{"id":"https://openalex.org/C105795698","wikidata":"https://www.wikidata.org/wiki/Q12483","display_name":"Statistics","level":1,"score":0.37130001187324524},{"id":"https://openalex.org/C126255220","wikidata":"https://www.wikidata.org/wiki/Q141495","display_name":"Mathematical optimization","level":1,"score":0.364300012588501},{"id":"https://openalex.org/C65660741","wikidata":"https://www.wikidata.org/wiki/Q3952743","display_name":"Score","level":2,"score":0.35679998993873596},{"id":"https://openalex.org/C2776029896","wikidata":"https://www.wikidata.org/wiki/Q3935810","display_name":"Relaxation (psychology)","level":2,"score":0.3513999879360199},{"id":"https://openalex.org/C186215838","wikidata":"https://www.wikidata.org/wiki/Q772232","display_name":"Conditional expectation","level":2,"score":0.34790000319480896},{"id":"https://openalex.org/C163716315","wikidata":"https://www.wikidata.org/wiki/Q901177","display_name":"Gaussian","level":2,"score":0.32249999046325684},{"id":"https://openalex.org/C21430997","wikidata":"https://www.wikidata.org/wiki/Q5159279","display_name":"Conditional variance","level":4,"score":0.32170000672340393},{"id":"https://openalex.org/C196083921","wikidata":"https://www.wikidata.org/wiki/Q7915758","display_name":"Variance (accounting)","level":2,"score":0.3188999891281128},{"id":"https://openalex.org/C111335779","wikidata":"https://www.wikidata.org/wiki/Q3454686","display_name":"Reduction (mathematics)","level":2,"score":0.3160000145435333},{"id":"https://openalex.org/C164172150","wikidata":"https://www.wikidata.org/wiki/Q1782585","display_name":"Consistent estimator","level":4,"score":0.31209999322891235},{"id":"https://openalex.org/C14036430","wikidata":"https://www.wikidata.org/wiki/Q3736076","display_name":"Function (biology)","level":2,"score":0.30970001220703125},{"id":"https://openalex.org/C159886148","wikidata":"https://www.wikidata.org/wiki/Q176645","display_name":"Markov process","level":2,"score":0.3041999936103821},{"id":"https://openalex.org/C39927690","wikidata":"https://www.wikidata.org/wiki/Q11197","display_name":"Logarithm","level":2,"score":0.29899999499320984},{"id":"https://openalex.org/C202615002","wikidata":"https://www.wikidata.org/wiki/Q783507","display_name":"Differentiable function","level":2,"score":0.28769999742507935},{"id":"https://openalex.org/C2778112365","wikidata":"https://www.wikidata.org/wiki/Q3511065","display_name":"Sequence (biology)","level":2,"score":0.2802000045776367},{"id":"https://openalex.org/C77553402","wikidata":"https://www.wikidata.org/wiki/Q13222579","display_name":"Upper and lower bounds","level":2,"score":0.2590000033378601},{"id":"https://openalex.org/C164660894","wikidata":"https://www.wikidata.org/wiki/Q2037833","display_name":"Piecewise","level":2,"score":0.257999986410141},{"id":"https://openalex.org/C971699","wikidata":"https://www.wikidata.org/wiki/Q1132714","display_name":"Delta method","level":3,"score":0.25619998574256897}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:oai:arXiv.org:2610.24188","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2610.24188","pdf_url":"https://arxiv.org/pdf/2610.24188","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,"listed_in":[],"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":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},{"id":"doi:10.48550/arxiv.2610.24188","doi":"https://doi.org/10.48550/arxiv.2610.24188","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2610.24188","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,"listed_in":[],"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-nc-sa","license_id":"https://openalex.org/licenses/cc-by-nc-sa","version":null,"is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"Preprint"}],"best_oa_location":{"id":"pmh:oai:arXiv.org:2610.24188","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2610.24188","pdf_url":"https://arxiv.org/pdf/2610.24188","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,"listed_in":[],"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":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},"sustainable_development_goals":[],"study_designs":[],"awards":[{"id":"https://openalex.org/G5971319010","display_name":"Self-Supervised NLP: From Text to Knowledge and Back","funder_award_id":"211379","funder_id":"https://openalex.org/F4320320924","funder_display_name":"Schweizerischer Nationalfonds zur F\u00f6rderung der Wissenschaftlichen Forschung"}],"funders":[{"id":"https://openalex.org/F4320320924","display_name":"Schweizerischer Nationalfonds zur F\u00f6rderung der Wissenschaftlichen Forschung","ror":"https://ror.org/00yjd3n13"}],"has_content":{"grobid_xml":true,"pdf":true},"content_urls":{"pdf":"https://content.openalex.org/works/W7223046115.pdf","grobid_xml":"https://content.openalex.org/works/W7223046115.grobid-xml"},"referenced_works_count":18,"referenced_works":["https://openalex.org/W1191599655","https://openalex.org/W1646152356","https://openalex.org/W1959608418","https://openalex.org/W2242818861","https://openalex.org/W2548228487","https://openalex.org/W2560592986","https://openalex.org/W2736601468","https://openalex.org/W2786614119","https://openalex.org/W2904157920","https://openalex.org/W2909637611","https://openalex.org/W2992977009","https://openalex.org/W3092500904","https://openalex.org/W3098134815","https://openalex.org/W3099014706","https://openalex.org/W4391631327","https://openalex.org/W4391941820","https://openalex.org/W4409147637","https://openalex.org/W6948256671"],"related_works":["https://openalex.org/W4392428184","https://openalex.org/W4391409203","https://openalex.org/W2769219710","https://openalex.org/W4229458284","https://openalex.org/W7213529782","https://openalex.org/W2991706463","https://openalex.org/W4402659302","https://openalex.org/W4413680324","https://openalex.org/W7077916309","https://openalex.org/W2285432433"],"abstract_inverted_index":{"Policy":[0],"gradients":[1,58],"for":[2,185],"autoregressive":[3,182],"sequence":[4],"models":[5],"differentiate":[6],"each":[7],"token's":[8],"log-probability":[9],"while":[10,125],"holding":[11],"the":[12,43,47,66,95,98,102,114,138,165],"conditioning":[13],"history":[14],"fixed.":[15],"This":[16],"discards":[17],"information":[18],"about":[19],"how":[20],"earlier":[21],"decisions":[22],"shape":[23],"later":[24],"conditional":[25],"distributions,":[26],"creating":[27],"a":[28,51,85,90,168,173,177],"credit-assignment":[29],"bottleneck":[30],"when":[31,76],"rewards":[32],"are":[33],"sparse.":[34],"We":[35,163],"introduce":[36],"History-Reparameterized":[37],"(HR)":[38],"policy":[39,115],"gradients,":[40],"which":[41],"replace":[42],"frozen":[44],"prefix":[45],"in":[46,113,160],"score":[48],"function":[49],"with":[50,89,127,152,176],"differentiable":[52],"reparameterized":[53],"or":[54,118],"relaxed":[55],"history,":[56],"allowing":[57],"of":[59,97,140],"downstream":[60],"log-probabilities":[61],"to":[62],"propagate":[63],"backward":[64],"through":[65],"generated":[67],"trajectory.":[68],"The":[69,105],"resulting":[70],"history-pathwise":[71,133],"term":[72],"is":[73,108,142],"generally":[74],"biased":[75],"added":[77],"directly,":[78],"so":[79],"we":[80],"derive":[81],"Optimal":[82],"HR":[83],"(OHR),":[84],"centered":[86],"control-variate":[87],"estimator":[88],"closed-form":[91],"coefficient":[92],"that":[93,150],"minimizes":[94],"trace":[96],"gradient":[99],"variance":[100,106,148],"at":[101],"population":[103],"level.":[104],"reduction":[107,149],"governed":[109],"by":[110],"temporal":[111],"structure":[112],"dynamics:":[116],"random":[117],"weakly":[119],"structured":[120],"policies":[121,126],"provide":[122],"little":[123],"benefit,":[124],"learned":[128],"sequential":[129],"dependencies":[130],"yield":[131],"informative":[132],"corrections.":[134],"For":[135],"discrete":[136],"tokens,":[137],"choice":[139],"relaxation":[141,145],"crucial:":[143],"embedding-level":[144],"yields":[146,157],"substantial":[147],"grows":[151],"horizon,":[153],"whereas":[154],"straight-through":[155],"Gumbel-Softmax":[156],"essentially":[158],"none":[159],"our":[161],"diagnostics.":[162],"validate":[164],"method":[166],"on":[167],"Gaussian":[169],"linear":[170],"Markov":[171],"model,":[172],"sparse-reward":[174],"T-maze":[175],"recurrent":[178],"policy,":[179],"and":[180],"an":[181],"Transformer":[183],"model":[184],"small":[186],"Traveling":[187],"Salesman":[188],"Problem":[189],"(TSP)":[190],"instances.":[191]},"counts_by_year":[],"updated_date":"2026-11-07T11:25:07.637180","created_date":"2026-10-21T00:00:00"}