{"id":"https://openalex.org/W4393729321","doi":"https://doi.org/10.48550/arxiv.2404.00934","title":"ChatGLM-RLHF: Practices of Aligning Large Language Models with Human Feedback","display_name":"ChatGLM-RLHF: Practices of Aligning Large Language Models with Human Feedback","publication_year":2024,"publication_date":"2024-04-01","ids":{"openalex":"https://openalex.org/W4393729321","doi":"https://doi.org/10.48550/arxiv.2404.00934"},"language":"en","primary_location":{"id":"pmh:oai:arXiv.org:2404.00934","doi":null,"is_oa":true,"landing_page_url":"http://arxiv.org/abs/2404.00934","pdf_url":"https://arxiv.org/pdf/2404.00934","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","license_id":"https://openalex.org/licenses/cc-by","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/2404.00934","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5060102585","display_name":"Zhenyu Hou","orcid":"https://orcid.org/0000-0002-1624-2149","observed_orcids":["https://orcid.org/0000-0002-1624-2149"]},"institutions":[{"id":"https://openalex.org/I99065089","display_name":"Tsinghua University","ror":"https://ror.org/03cve4549","country_code":"CN","type":"education","lineage":["https://openalex.org/I99065089"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Hou, Zhenyu","raw_affiliation_strings":["Tsinghua University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Tsinghua University","institution_ids":["https://openalex.org/I99065089"]}]},{"author_position":"middle","author":{"id":null,"display_name":"Niu, Yilin","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I99065089","display_name":"Tsinghua University","ror":"https://ror.org/03cve4549","country_code":"CN","type":"education","lineage":["https://openalex.org/I99065089"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Niu, Yilin","raw_affiliation_strings":["Tsinghua University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Tsinghua University","institution_ids":["https://openalex.org/I99065089"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5018336656","display_name":"Zhengxiao Du","orcid":"https://orcid.org/0000-0002-8223-4147","observed_orcids":["https://orcid.org/0000-0002-8223-4147"]},"institutions":[{"id":"https://openalex.org/I99065089","display_name":"Tsinghua University","ror":"https://ror.org/03cve4549","country_code":"CN","type":"education","lineage":["https://openalex.org/I99065089"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Du, Zhengxiao","raw_affiliation_strings":["Tsinghua University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Tsinghua University","institution_ids":["https://openalex.org/I99065089"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5100431077","display_name":"Xiaohan Zhang","orcid":"https://orcid.org/0000-0003-1522-4042","observed_orcids":["https://orcid.org/0000-0003-1522-4042","https://orcid.org/0000-0001-6305-2463","https://orcid.org/0009-0000-4673-4964","https://orcid.org/0009-0004-6120-4813","https://orcid.org/0000-0002-4758-8070"]},"institutions":[{"id":"https://openalex.org/I99065089","display_name":"Tsinghua University","ror":"https://ror.org/03cve4549","country_code":"CN","type":"education","lineage":["https://openalex.org/I99065089"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Zhang, Xiaohan","raw_affiliation_strings":["Tsinghua University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Tsinghua University","institution_ids":["https://openalex.org/I99065089"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5102612370","display_name":"Xiao Liu","orcid":"https://orcid.org/0000-0002-4405-3839","observed_orcids":["https://orcid.org/0000-0002-4405-3839"]},"institutions":[{"id":"https://openalex.org/I99065089","display_name":"Tsinghua University","ror":"https://ror.org/03cve4549","country_code":"CN","type":"education","lineage":["https://openalex.org/I99065089"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Liu, Xiao","raw_affiliation_strings":["Tsinghua University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Tsinghua University","institution_ids":["https://openalex.org/I99065089"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5093104489","display_name":"Aohan Zeng","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I99065089","display_name":"Tsinghua University","ror":"https://ror.org/03cve4549","country_code":"CN","type":"education","lineage":["https://openalex.org/I99065089"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Zeng, Aohan","raw_affiliation_strings":["Tsinghua University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Tsinghua University","institution_ids":["https://openalex.org/I99065089"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5065248294","display_name":"Qinkai Zheng","orcid":"https://orcid.org/0000-0002-5391-9446","observed_orcids":["https://orcid.org/0000-0002-5391-9446"]},"institutions":[{"id":"https://openalex.org/I99065089","display_name":"Tsinghua University","ror":"https://ror.org/03cve4549","country_code":"CN","type":"education","lineage":["https://openalex.org/I99065089"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Zheng, Qinkai","raw_affiliation_strings":["Tsinghua University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Tsinghua University","institution_ids":["https://openalex.org/I99065089"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5044042138","display_name":"Minlie Huang","orcid":"https://orcid.org/0000-0001-7111-1849","observed_orcids":["https://orcid.org/0000-0001-7111-1849"]},"institutions":[{"id":"https://openalex.org/I99065089","display_name":"Tsinghua University","ror":"https://ror.org/03cve4549","country_code":"CN","type":"education","lineage":["https://openalex.org/I99065089"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Huang, Minlie","raw_affiliation_strings":["Tsinghua University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Tsinghua University","institution_ids":["https://openalex.org/I99065089"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5085094109","display_name":"Hongning Wang","orcid":"https://orcid.org/0000-0002-6524-9195","observed_orcids":["https://orcid.org/0000-0002-6524-9195"]},"institutions":[{"id":"https://openalex.org/I99065089","display_name":"Tsinghua University","ror":"https://ror.org/03cve4549","country_code":"CN","type":"education","lineage":["https://openalex.org/I99065089"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Wang, Hongning","raw_affiliation_strings":["Tsinghua University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Tsinghua University","institution_ids":["https://openalex.org/I99065089"]}]},{"author_position":"middle","author":{"id":null,"display_name":"Tang, Jie","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I99065089","display_name":"Tsinghua University","ror":"https://ror.org/03cve4549","country_code":"CN","type":"education","lineage":["https://openalex.org/I99065089"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Tang, Jie","raw_affiliation_strings":["Tsinghua University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Tsinghua University","institution_ids":["https://openalex.org/I99065089"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5052284218","display_name":"Yuxiao Dong","orcid":"https://orcid.org/0000-0002-6092-2002","observed_orcids":["https://orcid.org/0000-0002-6092-2002"]},"institutions":[{"id":"https://openalex.org/I99065089","display_name":"Tsinghua University","ror":"https://ror.org/03cve4549","country_code":"CN","type":"education","lineage":["https://openalex.org/I99065089"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Dong, Yuxiao","raw_affiliation_strings":["Tsinghua University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Tsinghua University","institution_ids":["https://openalex.org/I99065089"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":1,"corresponding_author_ids":[],"corresponding_institution_ids":["https://openalex.org/I99065089"],"apc_list":null,"apc_paid":null,"fwci":null,"has_fulltext":true,"cited_by_count":3,"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/T10028","display_name":"Topic Modeling","score":0.24950000643730164,"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/T10028","display_name":"Topic Modeling","score":0.24950000643730164,"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":0.2313999980688095,"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/T10462","display_name":"Reinforcement Learning in Robotics","score":0.16329999268054962,"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/chatglm","display_name":"ChatGLM","score":0.9620000123977661},{"id":"https://openalex.org/keywords/reinforcement-learning-from-human-feedback","display_name":"reinforcement learning from human feedback","score":0.8539999723434448},{"id":"https://openalex.org/keywords/large-language-model-alignment","display_name":"large language model alignment","score":0.9409999847412109},{"id":"https://openalex.org/keywords/model-parallelism","display_name":"model parallelism","score":0.8679999709129333},{"id":"https://openalex.org/keywords/catastrophic-forgetting","display_name":"catastrophic forgetting","score":0.9620000123977661},{"id":"https://openalex.org/keywords/regularization","display_name":"regularization","score":0.6060000061988831}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.5005214214324951},{"id":"https://openalex.org/C41895202","wikidata":"https://www.wikidata.org/wiki/Q8162","display_name":"Linguistics","level":1,"score":0.3669578433036804},{"id":"https://openalex.org/C138885662","wikidata":"https://www.wikidata.org/wiki/Q5891","display_name":"Philosophy","level":0,"score":0.1176736056804657}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:oai:arXiv.org:2404.00934","doi":null,"is_oa":true,"landing_page_url":"http://arxiv.org/abs/2404.00934","pdf_url":"https://arxiv.org/pdf/2404.00934","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","license_id":"https://openalex.org/licenses/cc-by","version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},{"id":"doi:10.48550/arxiv.2404.00934","doi":"https://doi.org/10.48550/arxiv.2404.00934","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2404.00934","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","license_id":"https://openalex.org/licenses/cc-by","version":null,"is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"Preprint"}],"best_oa_location":{"id":"pmh:oai:arXiv.org:2404.00934","doi":null,"is_oa":true,"landing_page_url":"http://arxiv.org/abs/2404.00934","pdf_url":"https://arxiv.org/pdf/2404.00934","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","license_id":"https://openalex.org/licenses/cc-by","version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},"sustainable_development_goals":[],"sustainable_development_goals_aurora":[],"study_designs":[],"awards":[],"funders":[],"has_content":{"grobid_xml":true,"pdf":true},"content_urls":{"pdf":"https://content.openalex.org/works/W4393729321.pdf","grobid_xml":"https://content.openalex.org/works/W4393729321.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":["https://openalex.org/W4402683794","https://openalex.org/W7131779452","https://openalex.org/W4407764293","https://openalex.org/W4365211596","https://openalex.org/W4415316359","https://openalex.org/W7162466061","https://openalex.org/W4409361541","https://openalex.org/W4400025402","https://openalex.org/W4393118723","https://openalex.org/W4403852125"],"abstract_inverted_index":{"ChatGLM":[0,9],"is":[1],"a":[2,25],"free-to-use":[3],"AI":[4],"service":[5],"powered":[6],"by":[7],"the":[8,21,47,53,56,60,65,81,119,155],"family":[10],"of":[11,49,55,62,67,124,146],"large":[12],"language":[13],"models":[14],"(LLMs).":[15],"In":[16],"this":[17],"paper,":[18],"we":[19,72],"present":[20],"ChatGLM-RLHF":[22,42,69,110],"pipeline":[23],"--":[24,33],"reinforcement":[26],"learning":[27],"from":[28],"human":[29,40,50,150],"feedback":[30],"(RLHF)":[31],"system":[32],"designed":[34],"to":[35,83,101,118],"enhance":[36],"ChatGLM's":[37],"alignment":[38,115,139],"with":[39,94,149],"preferences.":[41],"encompasses":[43],"three":[44],"major":[45],"components:":[46],"collection":[48],"preference":[51],"data,":[52],"training":[54],"reward":[57,85],"model,":[58],"and":[59,74,97,157],"optimization":[61],"policies.":[63],"Throughout":[64],"process":[66],"integrating":[68],"into":[70,154],"production,":[71],"encountered":[73],"addressed":[75],"several":[76],"unprecedented":[77],"challenges.":[78],"We":[79],"introduce":[80],"strategies":[82],"mitigate":[84],"variance":[86],"for":[87],"stabilized":[88],"large-scale":[89],"training,":[90],"implement":[91],"model":[92],"parallelism":[93],"fused":[95],"gradient-descent,":[96],"design":[98],"regularization":[99],"constraints":[100],"avoid":[102],"catastrophic":[103],"forgetting":[104],"in":[105,114,137,159],"LLMs.":[106],"Experiments":[107],"show":[108],"that":[109],"brings":[111],"significant":[112],"improvements":[113],"tasks":[116],"compared":[117],"supervised":[120],"fine-tuned":[121],"(SFT)":[122],"version":[123],"ChatGLM.":[125],"For":[126],"instance,":[127],"it":[128],"achieves":[129],"on":[130],"average":[131],"15\\%":[132],"more":[133],"wins":[134],"against":[135],"ChatGLM-SFT":[136],"Chinese":[138],"tasks.":[140],"The":[141],"work":[142],"presents":[143],"our":[144],"practices":[145],"aligning":[147],"LLMs":[148],"preferences,":[151],"offering":[152],"insights":[153],"challenges":[156],"solutions":[158],"RLHF":[160],"implementations.":[161]},"counts_by_year":[{"year":2025,"cited_by_count":1},{"year":2024,"cited_by_count":2}],"updated_date":"2026-10-13T10:08:12.033832","created_date":"2024-04-03T00:00:00"}
