{"id":"https://openalex.org/W7128696271","doi":"https://doi.org/10.48448/myp6-b390","title":"1057 - TAPO: Dynamic Teacher and Perturbed Answer Injection for Policy Optimization","display_name":"1057 - TAPO: Dynamic Teacher and Perturbed Answer Injection for Policy Optimization","publication_year":2026,"publication_date":"2026-01-07","ids":{"openalex":"https://openalex.org/W7128696271","doi":"https://doi.org/10.48448/myp6-b390"},"language":"en","primary_location":{"id":"doi:10.48448/myp6-b390","doi":"https://doi.org/10.48448/myp6-b390","is_oa":true,"landing_page_url":"https://doi.org/10.48448/myp6-b390","pdf_url":null,"source":{"id":"https://openalex.org/S7407053148","display_name":"Underline Science Inc.","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"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":"Audiovisual"},"volume":null,"type":"other","is_traditional":false,"indexed_in":["datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://doi.org/10.48448/myp6-b390","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5125741655","display_name":"Association for Artificial Intelligence 2026","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Association for Artificial Intelligence 2026","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5125706351","display_name":"Liruiqi .","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"., Liruiqi","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5028315438","display_name":"Peter B\u00fa\u0161","orcid":"https://orcid.org/0000-0002-1730-4559","observed_orcids":["https://orcid.org/0000-0002-1730-4559"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"B\u00fa\u0161, Peter","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5101251764","display_name":"Moquan Cheng","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Cheng, Moquan","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5125738390","display_name":"Zhiyong Dong","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Dong, Zhiyong","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5125719726","display_name":"Yusong Hu","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Hu, Yusong","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5125710804","display_name":"Maowei Jiang","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Jiang, Maowei","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5129703827","display_name":"Alan Liang","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Liang, Alan","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5125696783","display_name":"Quangao Liu","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Liu, Quangao","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5125721283","display_name":"Ruikai Liu","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Liu, Ruikai","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5125687471","display_name":"Yifan Wang","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wang, Yifan","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5125717974","display_name":"Zihang Wang","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wang, Zihang","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5125694604","display_name":"Qi Feng Wang","orcid":"https://orcid.org/0000-0002-2688-8306","observed_orcids":["https://orcid.org/0000-0002-2688-8306"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wang, Qi","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5125726529","display_name":"Yansong Xu","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Xu, Yansong","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5125714921","display_name":"Pengyu 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, Pengyu","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/A5125733652","display_name":"Chaoran Zhang","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhang, Chaoran","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":1,"corresponding_author_ids":[],"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/T10028","display_name":"Topic Modeling","score":0.42160001397132874,"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":0.42160001397132874,"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":0.26589998602867126,"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":0.08550000190734863,"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/large-language-models","display_name":"large language models","score":0.6769999861717224},{"id":"https://openalex.org/keywords/group-relative-policy-optimization","display_name":"Group Relative Policy Optimization","score":0.8730000257492065},{"id":"https://openalex.org/keywords/reward-shaping","display_name":"reward shaping","score":0.6669999957084656},{"id":"https://openalex.org/keywords/mathematical-reasoning","display_name":"mathematical reasoning","score":0.7300000190734863},{"id":"https://openalex.org/keywords/adversarial-examples","display_name":"adversarial examples","score":0.7549999952316284},{"id":"https://openalex.org/keywords/semantic-redundancy","display_name":"semantic redundancy","score":0.6359999775886536}],"concepts":[{"id":"https://openalex.org/C55439883","wikidata":"https://www.wikidata.org/wiki/Q360812","display_name":"Correctness","level":2,"score":0.7922999858856201},{"id":"https://openalex.org/C97541855","wikidata":"https://www.wikidata.org/wiki/Q830687","display_name":"Reinforcement learning","level":2,"score":0.6848000288009644},{"id":"https://openalex.org/C37736160","wikidata":"https://www.wikidata.org/wiki/Q1801315","display_name":"Adversarial system","level":2,"score":0.5978999733924866},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.5831000208854675},{"id":"https://openalex.org/C112972136","wikidata":"https://www.wikidata.org/wiki/Q7595718","display_name":"Stability (learning theory)","level":2,"score":0.5476999878883362},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.5041000247001648},{"id":"https://openalex.org/C12725497","wikidata":"https://www.wikidata.org/wiki/Q810247","display_name":"Baseline (sea)","level":2,"score":0.5037999749183655},{"id":"https://openalex.org/C2164484","wikidata":"https://www.wikidata.org/wiki/Q5170150","display_name":"Core (optical fiber)","level":2,"score":0.4648999869823456},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.33570000529289246},{"id":"https://openalex.org/C67203356","wikidata":"https://www.wikidata.org/wiki/Q1321905","display_name":"Reinforcement","level":2,"score":0.3147999942302704},{"id":"https://openalex.org/C61797465","wikidata":"https://www.wikidata.org/wiki/Q1188986","display_name":"Term (time)","level":2,"score":0.27140000462532043},{"id":"https://openalex.org/C137836250","wikidata":"https://www.wikidata.org/wiki/Q984063","display_name":"Optimization problem","level":2,"score":0.2551000118255615},{"id":"https://openalex.org/C175154964","wikidata":"https://www.wikidata.org/wiki/Q380077","display_name":"Task analysis","level":3,"score":0.250900000333786}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48448/myp6-b390","doi":"https://doi.org/10.48448/myp6-b390","is_oa":true,"landing_page_url":"https://doi.org/10.48448/myp6-b390","pdf_url":null,"source":{"id":"https://openalex.org/S7407053148","display_name":"Underline Science Inc.","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"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":"Audiovisual"}],"best_oa_location":{"id":"doi:10.48448/myp6-b390","doi":"https://doi.org/10.48448/myp6-b390","is_oa":true,"landing_page_url":"https://doi.org/10.48448/myp6-b390","pdf_url":null,"source":{"id":"https://openalex.org/S7407053148","display_name":"Underline Science Inc.","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"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":"Audiovisual"},"sustainable_development_goals":[{"score":0.4156000018119812,"id":"https://metadata.un.org/sdg/4","display_name":"Quality education"}],"study_designs":[],"awards":[],"funders":[],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":["https://openalex.org/W7138884082","https://openalex.org/W7139928139","https://openalex.org/W7116901512","https://openalex.org/W7155085266","https://openalex.org/W7147282947","https://openalex.org/W4416248792","https://openalex.org/W7128535360","https://openalex.org/W7168407245","https://openalex.org/W7127640081","https://openalex.org/W4415191969"],"abstract_inverted_index":{"\\begin{abstract}":[0],"Reinforcement":[1],"learning":[2,87],"(RL)":[3],"has":[4],"emerged":[5],"as":[6,23],"a":[7,85,153,195],"powerful":[8],"framework":[9,88],"to":[10,47,117,135],"improve":[11],"the":[12,63,146,180,191,202],"reasoning":[13,64,140,182],"performance":[14],"of":[15,184,204],"large":[16],"language":[17],"models":[18],"(LLMs),":[19],"with":[20,38,67],"approaches":[21],"such":[22],"Group":[24],"Relative":[25],"Policy":[26],"Optimization":[27],"(GRPO)":[28],"showing":[29],"promising":[30],"results.":[31],"However,":[32],"GRPO":[33,192],"and":[34,50,75,148,164],"its":[35],"variants":[36],"struggle":[37],"collapsed":[39,123],"groups":[40],"(i.e.,":[41],"all-correct":[42],"or":[43,114],"all-incorrect":[44],"completions),":[45],"leading":[46],"zero-variance":[48],"rewards":[49],"ineffective":[51],"gradient":[52,120],"signals.":[53],"Moreover,":[54],"focusing":[55],"solely":[56],"on":[57,161],"final":[58],"answer":[59,144],"correctness":[60,141],"while":[61],"ignoring":[62],"process,":[65],"along":[66],"rigid":[68],"length":[69,163],"penalties,":[70],"can":[71],"hinder":[72],"training":[73,98],"stability":[74],"output":[76],"quality.":[77],"To":[78],"address":[79],"these":[80],"issues,":[81],"we":[82],"introduce":[83],"TAPO,":[84],"reinforcement":[86],"that":[89,157,176],"enhances":[90],"optimization":[91],"signals":[92,121],"by":[93,194],"modifying":[94],"sampled":[95],"completions":[96,134],"within":[97],"groups.":[99],"TAPO":[100,177],"incorporates":[101],"three":[102],"core":[103],"techniques:":[104],"(1)":[105],"\\textit{Dynamic":[106],"Teacher":[107],"Injection":[108,128],"(DTI)},":[109],"which":[110,130],"selectively":[111],"injects":[112],"high-quality":[113],"adversarial":[115],"examples":[116],"restore":[118],"effective":[119],"in":[122],"groups;":[124],"(2)":[125],"\\textit{Perturbed":[126],"Answer":[127],"(PAI)},":[129],"makes":[131],"partially":[132],"correct":[133],"provide":[136],"contrastive":[137],"supervision":[138],"separating":[139],"but":[142],"wrong":[143],"from":[145],"trajectories;":[147],"(3)":[149],"\\textit{InfoLen-Aware":[150],"Reward":[151],"Shaping},":[152],"fine-grained":[154],"reward":[155],"strategy":[156],"penalizes":[158],"outputs":[159],"based":[160],"both":[162],"semantic":[165],"redundancy,":[166],"encouraging":[167],"concise":[168],"yet":[169],"informative":[170],"responses.":[171],"Extensive":[172],"experimental":[173],"results":[174],"demonstrate":[175],"significantly":[178],"improves":[179],"mathematical":[181],"capabilities":[183],"LLMs":[185],"across":[186],"multiple":[187],"challenging":[188],"benchmarks,":[189],"outperforming":[190],"baseline":[193],"substantial":[196],"margin.":[197],"Component-wise":[198],"ablations":[199],"further":[200],"validate":[201],"contribution":[203],"each":[205],"proposed":[206],"technique.":[207],"\\end{abstract}":[208]},"counts_by_year":[],"updated_date":"2026-10-28T10:33:24.094632","created_date":"2026-02-13T00:00:00"}