{"id":"https://openalex.org/W4405418995","doi":"https://doi.org/10.48550/arxiv.2408.11791","title":"Critique-out-Loud Reward Models","display_name":"Critique-out-Loud Reward Models","publication_year":2024,"publication_date":"2024-08-21","ids":{"openalex":"https://openalex.org/W4405418995","doi":"https://doi.org/10.48550/arxiv.2408.11791"},"language":"en","primary_location":{"id":"pmh:oai:arXiv.org:2408.11791","doi":null,"is_oa":true,"landing_page_url":"http://arxiv.org/abs/2408.11791","pdf_url":"https://arxiv.org/pdf/2408.11791","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":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},"volume":null,"type":"preprint","indexed_in":["arxiv","datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://arxiv.org/pdf/2408.11791","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5024750873","display_name":"Zachary Ankner","orcid":"https://orcid.org/0000-0001-8307-3456","observed_orcids":["https://orcid.org/0000-0001-8307-3456"]},"institutions":[{"id":"https://openalex.org/I63966007","display_name":"Massachusetts Institute of Technology","ror":"https://ror.org/042nb2s44","country_code":"US","type":"education","lineage":["https://openalex.org/I63966007"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Ankner, Zachary","raw_affiliation_strings":["MIT"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"MIT","institution_ids":["https://openalex.org/I63966007"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5036307641","display_name":"Mansheej Paul","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Paul, Mansheej","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":null,"display_name":"Cui, Brandon","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Cui, Brandon","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5115454653","display_name":"Jonathan D. Chang","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Chang, Jonathan D.","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5029637639","display_name":"Prithviraj Ammanabrolu","orcid":"https://orcid.org/0000-0003-1950-0454","observed_orcids":["https://orcid.org/0000-0003-1950-0454"]},"institutions":[{"id":"https://openalex.org/I36258959","display_name":"University of California San Diego","ror":"https://ror.org/0168r3w48","country_code":"US","type":"education","lineage":["https://openalex.org/I36258959"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Ammanabrolu, Prithviraj","raw_affiliation_strings":["University of California , San Diego"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"University of California , San Diego","institution_ids":["https://openalex.org/I36258959"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":2,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":null,"has_fulltext":true,"cited_by_count":28,"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":0.21960000693798065,"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":0.21960000693798065,"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/T10181","display_name":"Natural Language Processing Techniques","score":0.21809999644756317,"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/T10028","display_name":"Topic Modeling","score":0.20399999618530273,"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/reward-models","display_name":"reward models","score":0.7300000190734863},{"id":"https://openalex.org/keywords/reinforcement-learning-from-human-feedback","display_name":"reinforcement learning from human feedback","score":0.9139999747276306},{"id":"https://openalex.org/keywords/large-language-models","display_name":"large language models","score":0.7910000085830688},{"id":"https://openalex.org/keywords/llama-3","display_name":"Llama 3","score":0.6299999952316284}],"concepts":[{"id":"https://openalex.org/C15744967","wikidata":"https://www.wikidata.org/wiki/Q9418","display_name":"Psychology","level":0,"score":0.5367480516433716}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:oai:arXiv.org:2408.11791","doi":null,"is_oa":true,"landing_page_url":"http://arxiv.org/abs/2408.11791","pdf_url":"https://arxiv.org/pdf/2408.11791","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":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},{"id":"doi:10.48550/arxiv.2408.11791","doi":"https://doi.org/10.48550/arxiv.2408.11791","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2408.11791","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":null,"license_id":null,"version":null,"is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"Preprint"}],"best_oa_location":{"id":"pmh:oai:arXiv.org:2408.11791","doi":null,"is_oa":true,"landing_page_url":"http://arxiv.org/abs/2408.11791","pdf_url":"https://arxiv.org/pdf/2408.11791","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":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},"sustainable_development_goals":[],"study_designs":[],"awards":[],"funders":[],"has_content":{"grobid_xml":true,"pdf":true},"content_urls":{"pdf":"https://content.openalex.org/works/W4405418995.pdf","grobid_xml":"https://content.openalex.org/works/W4405418995.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":["https://openalex.org/W4404988001","https://openalex.org/W4404303906","https://openalex.org/W4393063868","https://openalex.org/W4411120032","https://openalex.org/W7131433391","https://openalex.org/W7102415536","https://openalex.org/W4404782113","https://openalex.org/W7134860561","https://openalex.org/W7134031930","https://openalex.org/W4417299315"],"abstract_inverted_index":{"Traditionally,":[0],"reward":[1,35,64,79,82,104,117,129,132,157,190,197],"models":[2,36,65,83,118,130,133,153,158,191],"used":[3,99,170],"for":[4,105,119,147,164,175,196],"reinforcement":[5],"learning":[6],"from":[7],"human":[8],"feedback":[9],"(RLHF)":[10],"are":[11],"trained":[12],"to":[13,66,100,127,160,181],"directly":[14],"predict":[15,101],"preference":[16,49,136],"scores":[17],"without":[18],"leveraging":[19],"the":[20,24,32,43,60,70,93,106,109,113,148,172,183],"generation":[21],"capabilities":[22,33,187],"of":[23,34,45,72,92,108,115,188],"underlying":[25],"large":[26],"language":[27,90],"model":[28,174],"(LLM).":[29],"This":[30],"limits":[31],"as":[37,171],"they":[38],"must":[39,51],"reason":[40,67],"implicitly":[41],"about":[42,69],"quality":[44,71,107],"a":[46,55,73,88,102,161],"response,":[47,74],"i.e.,":[48],"modeling":[50],"be":[52],"performed":[53],"in":[54],"single":[56],"forward":[57],"pass":[58],"through":[59],"model.":[61],"To":[62],"enable":[63],"explicitly":[68],"we":[75,178],"introduce":[76],"Critique-out-Loud":[77],"(CLoud)":[78],"models.":[80],"CLoud":[81,116,131,156,189],"operate":[84],"by":[85,141,192],"first":[86],"generating":[87],"natural":[89],"critique":[91],"assistant's":[94],"response":[95],"that":[96],"is":[97],"then":[98],"scalar":[103],"response.":[110],"We":[111],"demonstrate":[112],"success":[114],"both":[120],"Llama-3-8B":[121],"and":[122,143,150],"70B":[123,151],"base":[124,152],"models:":[125],"compared":[126],"classic":[128],"improve":[134],"pairwise":[135],"classification":[137],"accuracy":[138],"on":[139,167],"RewardBench":[140],"4.65":[142],"5.84":[144],"percentage":[145],"points":[146],"8B":[149],"respectively.":[154],"Furthermore,":[155],"lead":[159],"Pareto":[162],"improvement":[163],"win":[165],"rate":[166],"ArenaHard":[168],"when":[169],"scoring":[173],"Best-of-N.":[176],"Finally,":[177],"explore":[179],"how":[180],"exploit":[182],"dynamic":[184],"inference":[185],"compute":[186],"performing":[193],"self-consistency":[194],"decoding":[195],"prediction.":[198]},"counts_by_year":[{"year":2026,"cited_by_count":12},{"year":2025,"cited_by_count":16}],"updated_date":"2026-11-04T10:54:05.112534","created_date":"2024-12-16T00:00:00"}