{"id":"https://openalex.org/W7165139933","doi":"https://doi.org/10.48550/arxiv.2606.18844","title":"Learning from Your Own Mistakes: Constructing Learnable Micro-Reflective Trajectories for Self-Distillation","display_name":"Learning from Your Own Mistakes: Constructing Learnable Micro-Reflective Trajectories for Self-Distillation","publication_year":2026,"publication_date":"2026-06-17","ids":{"openalex":"https://openalex.org/W7165139933","doi":"https://doi.org/10.48550/arxiv.2606.18844"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2606.18844","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2606.18844","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":null,"issn":null,"is_oa":true,"is_in_doaj":false,"is_core":false,"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"},"type":"preprint","indexed_in":["datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://doi.org/10.48550/arxiv.2606.18844","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5011086655","display_name":"Z Zhangqin Huang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Huang, Zhilin","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5076692827","display_name":"He Gao","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Gao, Hang","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5136491825","display_name":"Ziqiang Dong","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Dong, Ziqiang","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5138905268","display_name":"Yuan Chen","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Chen, Yuan","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5125075910","display_name":"Yifeng Luo","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Luo, Yifeng","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5034643682","display_name":"Chujun Qin","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Qin, Chujun","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5138923563","display_name":"Jingyi Wang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wang, Jingyi","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5138920800","display_name":"Yang Yang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Yang, Yang","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5138850548","display_name":"Guanjun Jiang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Jiang, Guanjun","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":0,"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":0.5735999941825867,"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":0.5735999941825867,"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/T11714","display_name":"Multimodal Machine Learning Applications","score":0.08269999921321869,"subfield":{"id":"https://openalex.org/subfields/1707","display_name":"Computer Vision and Pattern Recognition"},"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/T11307","display_name":"Domain Adaptation and Few-Shot Learning","score":0.06599999964237213,"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/trajectory","display_name":"Trajectory","score":0.5317000150680542},{"id":"https://openalex.org/keywords/point","display_name":"Point (geometry)","score":0.4747999906539917},{"id":"https://openalex.org/keywords/divergence","display_name":"Divergence (linguistics)","score":0.41769999265670776},{"id":"https://openalex.org/keywords/selection","display_name":"Selection (genetic algorithm)","score":0.41119998693466187},{"id":"https://openalex.org/keywords/boundary","display_name":"Boundary (topology)","score":0.3781000077724457},{"id":"https://openalex.org/keywords/construct","display_name":"Construct (python library)","score":0.3749000132083893},{"id":"https://openalex.org/keywords/sampling","display_name":"Sampling (signal processing)","score":0.34380000829696655},{"id":"https://openalex.org/keywords/grammaticality","display_name":"Grammaticality","score":0.30320000648498535},{"id":"https://openalex.org/keywords/gradient-descent","display_name":"Gradient descent","score":0.2928999960422516}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.738099992275238},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.5950000286102295},{"id":"https://openalex.org/C13662910","wikidata":"https://www.wikidata.org/wiki/Q193139","display_name":"Trajectory","level":2,"score":0.5317000150680542},{"id":"https://openalex.org/C28719098","wikidata":"https://www.wikidata.org/wiki/Q44946","display_name":"Point (geometry)","level":2,"score":0.4747999906539917},{"id":"https://openalex.org/C207390915","wikidata":"https://www.wikidata.org/wiki/Q1230525","display_name":"Divergence (linguistics)","level":2,"score":0.41769999265670776},{"id":"https://openalex.org/C81917197","wikidata":"https://www.wikidata.org/wiki/Q628760","display_name":"Selection (genetic algorithm)","level":2,"score":0.41119998693466187},{"id":"https://openalex.org/C62354387","wikidata":"https://www.wikidata.org/wiki/Q875399","display_name":"Boundary (topology)","level":2,"score":0.3781000077724457},{"id":"https://openalex.org/C2780801425","wikidata":"https://www.wikidata.org/wiki/Q5164392","display_name":"Construct (python library)","level":2,"score":0.3749000132083893},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.3646000027656555},{"id":"https://openalex.org/C140779682","wikidata":"https://www.wikidata.org/wiki/Q210868","display_name":"Sampling (signal processing)","level":3,"score":0.34380000829696655},{"id":"https://openalex.org/C2779525943","wikidata":"https://www.wikidata.org/wiki/Q1187300","display_name":"Grammaticality","level":3,"score":0.30320000648498535},{"id":"https://openalex.org/C31972630","wikidata":"https://www.wikidata.org/wiki/Q844240","display_name":"Computer vision","level":1,"score":0.2948000133037567},{"id":"https://openalex.org/C153258448","wikidata":"https://www.wikidata.org/wiki/Q1199743","display_name":"Gradient descent","level":3,"score":0.2928999960422516},{"id":"https://openalex.org/C20162079","wikidata":"https://www.wikidata.org/wiki/Q1151406","display_name":"Case-based reasoning","level":2,"score":0.29260000586509705},{"id":"https://openalex.org/C118317068","wikidata":"https://www.wikidata.org/wiki/Q2100760","display_name":"Point distribution model","level":2,"score":0.2896000146865845},{"id":"https://openalex.org/C2777212361","wikidata":"https://www.wikidata.org/wiki/Q5127848","display_name":"Class (philosophy)","level":2,"score":0.28769999742507935},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.28679999709129333},{"id":"https://openalex.org/C2779843651","wikidata":"https://www.wikidata.org/wiki/Q7390335","display_name":"SIGNAL (programming language)","level":2,"score":0.28290000557899475},{"id":"https://openalex.org/C2778334786","wikidata":"https://www.wikidata.org/wiki/Q1586270","display_name":"Variation (astronomy)","level":2,"score":0.2818000018596649},{"id":"https://openalex.org/C103278499","wikidata":"https://www.wikidata.org/wiki/Q254465","display_name":"Similarity (geometry)","level":3,"score":0.2808000147342682},{"id":"https://openalex.org/C42023084","wikidata":"https://www.wikidata.org/wiki/Q5249231","display_name":"Decision boundary","level":3,"score":0.28029999136924744},{"id":"https://openalex.org/C22367795","wikidata":"https://www.wikidata.org/wiki/Q7625208","display_name":"Structured prediction","level":2,"score":0.27889999747276306},{"id":"https://openalex.org/C110121322","wikidata":"https://www.wikidata.org/wiki/Q865811","display_name":"Distribution (mathematics)","level":2,"score":0.275299996137619},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.2727999985218048},{"id":"https://openalex.org/C2779795794","wikidata":"https://www.wikidata.org/wiki/Q7315343","display_name":"Reset (finance)","level":2,"score":0.2662999927997589},{"id":"https://openalex.org/C40506919","wikidata":"https://www.wikidata.org/wiki/Q7452469","display_name":"Sequence learning","level":2,"score":0.25780001282691956},{"id":"https://openalex.org/C51632099","wikidata":"https://www.wikidata.org/wiki/Q3985153","display_name":"Training set","level":2,"score":0.25589999556541443}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2606.18844","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2606.18844","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":null,"issn":null,"is_oa":true,"is_in_doaj":false,"is_core":false,"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":null,"raw_source_name":null,"raw_type":"Preprint"}],"best_oa_location":{"id":"doi:10.48550/arxiv.2606.18844","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2606.18844","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":null,"issn":null,"is_oa":true,"is_in_doaj":false,"is_core":false,"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"},"sustainable_development_goals":[],"awards":[],"funders":[],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Self-distillation":[0],"improves":[1],"reasoning":[2,77,132,146,243],"in":[3,162],"large":[4],"language":[5],"models":[6],"by":[7,148,186],"using":[8],"the":[9,45,58,102,111,129,135,153,163,169,173,182,199,229],"model's":[10,46,130,174,200],"own":[11,165],"rollouts":[12,109],"as":[13],"training":[14,125,233],"signal,":[15],"typically":[16],"through":[17],"implicit":[18,92],"logit-level":[19],"alignment":[20,94,184],"that":[21,71,127,221,238],"minimizes":[22],"KL":[23],"divergence":[24],"toward":[25],"a":[26,63,141,149,178],"privileged":[27,64],"target":[28],"distribution.":[29],"However,":[30],"because":[31],"this":[32,80,117],"supervision":[33],"is":[34,160],"generated":[35],"via":[36],"uncontrolled":[37],"sampling,":[38],"it":[39],"provides":[40],"no":[41],"diagnostic":[42],"insight":[43],"into":[44],"specific":[47],"errors":[48],"or":[49],"corrective":[50,170],"guidance":[51],"for":[52],"its":[53,76],"individual":[54],"failure":[55],"patterns.":[56],"Consequently,":[57],"model":[59,103],"learns":[60],"to":[61,95,110,120,134,177,207],"imitate":[62],"distribution":[65,176],"rather":[66],"than":[67,181],"receiving":[68],"fine-grained":[69],"corrections":[70],"pinpoint":[72],"where":[73],"and":[74,107,114,144,167,203,217,244],"why":[75],"fails.":[78],"In":[79],"paper,":[81],"we":[82],"propose":[83],"Trajectory-Augmented":[84],"Policy":[85],"Optimization":[86],"(TAPO),":[87],"which":[88],"advances":[89],"self-distillation":[90],"from":[91,152],"distributional":[93],"explicit":[96],"trajectory":[97,159],"construction.":[98],"During":[99],"RL":[100],"training,":[101],"produces":[104],"both":[105,241],"correct":[106,150],"incorrect":[108],"same":[112,154,230],"query,":[113],"TAPO":[115,193,222,239],"leverages":[116],"contrastive":[118],"structure":[119],"construct":[121],"micro-reflective":[122],"corrections,":[123],"new":[124],"trajectories":[126],"retain":[128],"erroneous":[131],"up":[133],"point":[136],"of":[137,232],"failure,":[138],"then":[139],"insert":[140],"natural-language":[142],"diagnosis":[143],"corrected":[145],"guided":[147],"reference":[151],"sampling":[155],"group.":[156],"Since":[157],"each":[158],"anchored":[161],"learner's":[164],"prefix":[166],"solutions,":[168],"signal":[171],"preserves":[172],"on-policy":[175],"greater":[179],"extent":[180],"position-wise":[183],"imposed":[185],"KL-based":[187],"methods.":[188],"To":[189],"integrate":[190],"these":[191],"trajectories,":[192],"introduces":[194],"difficulty-aware":[195],"candidate":[196],"selection":[197],"at":[198],"capability":[201],"boundary":[202],"decoupled":[204],"advantage":[205],"estimation":[206],"prevent":[208],"gradient":[209],"contamination.":[210],"Experiments":[211],"on":[212],"AIME":[213,215],"2024,":[214],"2025,":[216],"HMMT":[218],"2025":[219],"show":[220],"achieves":[223],"consistent":[224],"improvements":[225],"over":[226],"GRPO":[227],"under":[228],"number":[231],"steps.":[234],"Further":[235],"analysis":[236],"demonstrates":[237],"strengthens":[240],"first-pass":[242],"error-correction":[245],"effectiveness.":[246]},"counts_by_year":[],"updated_date":"2026-07-01T06:00:48.157686","created_date":"2026-06-19T00:00:00"}