{"id":"https://openalex.org/W4417005376","doi":"https://doi.org/10.48550/arxiv.2512.02389","title":"Synthetic Error Injection Fails to Elicit Self-Correction In Language Models","display_name":"Synthetic Error Injection Fails to Elicit Self-Correction In Language Models","publication_year":2025,"publication_date":"2025-12-02","ids":{"openalex":"https://openalex.org/W4417005376","doi":"https://doi.org/10.48550/arxiv.2512.02389"},"language":"en","primary_location":{"id":"pmh:oai:arXiv.org:2512.02389","doi":null,"is_oa":true,"landing_page_url":"http://arxiv.org/abs/2512.02389","pdf_url":"https://arxiv.org/pdf/2512.02389","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","is_traditional":true,"indexed_in":["arxiv","datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://arxiv.org/pdf/2512.02389","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5104192939","display_name":"David Wu","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I95457486","display_name":"University of California, Berkeley","ror":"https://ror.org/01an7q238","country_code":"US","type":"education","lineage":["https://openalex.org/I95457486"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Wu, David X.","raw_affiliation_strings":["Department of EECS , UC Berkeley , Berkeley , CA , USA"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Department of EECS , UC Berkeley , Berkeley , CA , USA","institution_ids":["https://openalex.org/I95457486"]}]},{"author_position":"middle","author":{"id":null,"display_name":"Kapur, Shreyas","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I95457486","display_name":"University of California, Berkeley","ror":"https://ror.org/01an7q238","country_code":"US","type":"education","lineage":["https://openalex.org/I95457486"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Kapur, Shreyas","raw_affiliation_strings":["Department of EECS , UC Berkeley , Berkeley , CA , USA"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Department of EECS , UC Berkeley , Berkeley , CA , USA","institution_ids":["https://openalex.org/I95457486"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5079317491","display_name":"Anant Sahai","orcid":"https://orcid.org/0000-0001-9263-7719","observed_orcids":["https://orcid.org/0000-0001-9263-7719"]},"institutions":[{"id":"https://openalex.org/I95457486","display_name":"University of California, Berkeley","ror":"https://ror.org/01an7q238","country_code":"US","type":"education","lineage":["https://openalex.org/I95457486"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Sahai, Anant","raw_affiliation_strings":["Department of EECS , UC Berkeley , Berkeley , CA , USA"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Department of EECS , UC Berkeley , Berkeley , CA , USA","institution_ids":["https://openalex.org/I95457486"]}]},{"author_position":"last","author":{"id":null,"display_name":"Russell, Stuart","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I95457486","display_name":"University of California, Berkeley","ror":"https://ror.org/01an7q238","country_code":"US","type":"education","lineage":["https://openalex.org/I95457486"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Russell, Stuart","raw_affiliation_strings":["Department of EECS , UC Berkeley , Berkeley , CA , USA"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Department of EECS , UC Berkeley , Berkeley , CA , USA","institution_ids":["https://openalex.org/I95457486"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":1,"corresponding_author_ids":[],"corresponding_institution_ids":["https://openalex.org/I95457486"],"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/T10028","display_name":"Topic Modeling","score":0.6230000257492065,"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":0.6230000257492065,"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":0.24289999902248383,"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/T11689","display_name":"Adversarial Robustness in Machine Learning","score":0.018799999728798866,"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/self-correction","display_name":"self-correction","score":0.8759999871253967},{"id":"https://openalex.org/keywords/large-language-models","display_name":"large language models","score":0.9490000009536743},{"id":"https://openalex.org/keywords/reasoning-chains","display_name":"reasoning chains","score":0.7870000004768372},{"id":"https://openalex.org/keywords/error-masking","display_name":"error masking","score":0.3790000081062317},{"id":"https://openalex.org/keywords/distribution-shift","display_name":"distribution shift","score":0.8569999933242798},{"id":"https://openalex.org/keywords/supervised-fine-tuning","display_name":"supervised fine-tuning","score":0.6769999861717224},{"id":"https://openalex.org/keywords/reinforcement-learning-for-reasoning","display_name":"reinforcement learning for reasoning","score":0.5550000071525574}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.718500018119812},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.6446999907493591},{"id":"https://openalex.org/C97541855","wikidata":"https://www.wikidata.org/wiki/Q830687","display_name":"Reinforcement learning","level":2,"score":0.6003999710083008},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.5471000075340271},{"id":"https://openalex.org/C160920958","wikidata":"https://www.wikidata.org/wiki/Q7662746","display_name":"Synthetic data","level":2,"score":0.5392000079154968},{"id":"https://openalex.org/C2780586882","wikidata":"https://www.wikidata.org/wiki/Q7520643","display_name":"Simple (philosophy)","level":2,"score":0.48820000886917114},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.45980000495910645},{"id":"https://openalex.org/C2983448237","wikidata":"https://www.wikidata.org/wiki/Q1078276","display_name":"Language understanding","level":2,"score":0.3465000092983246},{"id":"https://openalex.org/C195324797","wikidata":"https://www.wikidata.org/wiki/Q33742","display_name":"Natural language","level":2,"score":0.3407999873161316},{"id":"https://openalex.org/C74672266","wikidata":"https://www.wikidata.org/wiki/Q815859","display_name":"Language acquisition","level":2,"score":0.31040000915527344},{"id":"https://openalex.org/C51632099","wikidata":"https://www.wikidata.org/wiki/Q3985153","display_name":"Training set","level":2,"score":0.3100000023841858},{"id":"https://openalex.org/C67203356","wikidata":"https://www.wikidata.org/wiki/Q1321905","display_name":"Reinforcement","level":2,"score":0.30550000071525574},{"id":"https://openalex.org/C204321447","wikidata":"https://www.wikidata.org/wiki/Q30642","display_name":"Natural language processing","level":1,"score":0.3001999855041504},{"id":"https://openalex.org/C167085575","wikidata":"https://www.wikidata.org/wiki/Q6803654","display_name":"Mean squared prediction error","level":2,"score":0.27799999713897705},{"id":"https://openalex.org/C103088060","wikidata":"https://www.wikidata.org/wiki/Q1062839","display_name":"Error detection and correction","level":2,"score":0.2711000144481659},{"id":"https://openalex.org/C94922259","wikidata":"https://www.wikidata.org/wiki/Q33215","display_name":"Constructed language","level":2,"score":0.2669999897480011},{"id":"https://openalex.org/C136389625","wikidata":"https://www.wikidata.org/wiki/Q334384","display_name":"Supervised learning","level":3,"score":0.2655999958515167},{"id":"https://openalex.org/C149441793","wikidata":"https://www.wikidata.org/wiki/Q200726","display_name":"Probability distribution","level":2,"score":0.2563000023365021}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:oai:arXiv.org:2512.02389","doi":null,"is_oa":true,"landing_page_url":"http://arxiv.org/abs/2512.02389","pdf_url":"https://arxiv.org/pdf/2512.02389","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.2512.02389","doi":"https://doi.org/10.48550/arxiv.2512.02389","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2512.02389","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:2512.02389","doi":null,"is_oa":true,"landing_page_url":"http://arxiv.org/abs/2512.02389","pdf_url":"https://arxiv.org/pdf/2512.02389","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":[],"study_designs":[],"awards":[{"id":"https://openalex.org/G6852334964","display_name":"Graduate Research Fellowship Program (GRFP)","funder_award_id":"2146752","funder_id":"https://openalex.org/F4320306076","funder_display_name":"National Science Foundation"}],"funders":[{"id":"https://openalex.org/F4320306076","display_name":"National Science Foundation","ror":"https://ror.org/021nxhr62"}],"has_content":{"pdf":true,"grobid_xml":true},"content_urls":{"pdf":"https://content.openalex.org/works/W4417005376.pdf","grobid_xml":"https://content.openalex.org/works/W4417005376.grobid-xml"},"referenced_works_count":9,"referenced_works":["https://openalex.org/W3134642945","https://openalex.org/W4313483736","https://openalex.org/W4385574274","https://openalex.org/W4403885103","https://openalex.org/W4404987838","https://openalex.org/W4405766390","https://openalex.org/W4414281281","https://openalex.org/W6948369276","https://openalex.org/W7077044503"],"related_works":["https://openalex.org/W4414959296","https://openalex.org/W4403885103","https://openalex.org/W7127318121","https://openalex.org/W4417537343","https://openalex.org/W7221550777","https://openalex.org/W7123581660","https://openalex.org/W4416149163","https://openalex.org/W4402669701","https://openalex.org/W7163708183","https://openalex.org/W7221071266"],"abstract_inverted_index":{"Reinforcement":[0],"learning":[1,37,144],"has":[2],"become":[3],"the":[4,61,70,96,105,111,122,126],"dominant":[5],"paradigm":[6],"for":[7,150],"eliciting":[8,151],"reasoning":[9,55],"and":[10,31,59,65],"self-correction":[11,44],"capabilities":[12,124],"in":[13,46],"large":[14],"language":[15,47],"models,":[16],"but":[17],"its":[18,99],"computational":[19],"expense":[20],"motivates":[21],"exploration":[22],"of":[23,73,114,125,134],"alternatives.":[24],"Inspired":[25],"by":[26],"techniques":[27],"from":[28],"autonomous":[29],"driving":[30],"robotics,":[32],"we":[33,76],"investigate":[34],"whether":[35],"supervised":[36],"with":[38,130],"synthetic":[39,88,115,132],"error":[40],"injection":[41],"can":[42],"induce":[43],"abilities":[45],"models.":[48,92],"Our":[49,137],"approach":[50],"inserts":[51],"artificial":[52],"errors":[53,116,119],"into":[54],"chains,":[56],"masks":[57],"them,":[58],"supervises":[60],"model":[62,97],"to":[63,81,117],"recognize":[64],"correct":[66],"these":[67],"mistakes.":[68],"Despite":[69],"intuitive":[71],"appeal":[72],"this":[74],"method,":[75],"find":[77,109],"that":[78,110],"it":[79,102],"fails":[80],"significantly":[82,120],"improve":[83],"performance":[84],"even":[85,94,129],"on":[86],"simple":[87],"tasks":[89],"across":[90],"multiple":[91],"Moreover,":[93],"when":[95],"catches":[98],"own":[100],"error,":[101],"often":[103],"parrots":[104],"original":[106],"mistake.":[107],"We":[108],"distribution":[112],"shift":[113],"on-policy":[118,135,142],"degrades":[121],"error-correction":[123],"fine-tuned":[127],"model,":[128],"good":[131],"coverage":[133],"errors.":[136],"results":[138],"help":[139],"explain":[140],"why":[141],"reinforcement":[143],"methods":[145],"have":[146],"proven":[147],"uniquely":[148],"effective":[149],"self-correction.":[152]},"counts_by_year":[],"updated_date":"2026-11-10T11:39:43.021295","created_date":"2025-12-04T00:00:00"}