{"id":"https://openalex.org/W7201834752","doi":"https://doi.org/10.48550/arxiv.2608.05810","title":"When Self-Evolution Backfires: Pre-Commit Gating against Skill Contamination in LLM Agents","display_name":"When Self-Evolution Backfires: Pre-Commit Gating against Skill Contamination in LLM Agents","publication_year":2026,"publication_date":"2026-08-06","ids":{"openalex":"https://openalex.org/W7201834752","doi":"https://doi.org/10.48550/arxiv.2608.05810"},"language":"en","primary_location":{"id":"pmh:oai:arXiv.org:2608.05810","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2608.05810","pdf_url":"https://arxiv.org/pdf/2608.05810","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":"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/2608.05810","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5127120639","display_name":"Linfang Shang","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Shang, Linfang","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5127082154","display_name":"Ming Xu","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Xu, Ming","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5127310753","display_name":"Yiding Sun","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Sun, Yiding","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5126255369","display_name":"Tianle Xia","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Xia, Tianle","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5126243813","display_name":"Lingxiang Hu","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Hu, Lingxiang","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5145934995","display_name":"Lan Xu","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Xu, Lan","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5145976731","display_name":"Ning Zheng","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zheng, Ning","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":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/T11975","display_name":"Evolutionary Algorithms and Applications","score":0.7202000021934509,"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/T11975","display_name":"Evolutionary Algorithms and Applications","score":0.7202000021934509,"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/T10906","display_name":"AI-based Problem Solving and Planning","score":0.10350000113248825,"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":0.023099999874830246,"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/phase-transitions","display_name":"phase transitions","score":0.6700000166893005}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6165000200271606},{"id":"https://openalex.org/C2776436953","wikidata":"https://www.wikidata.org/wiki/Q5163215","display_name":"Consistency (knowledge bases)","level":2,"score":0.5464000105857849},{"id":"https://openalex.org/C185798385","wikidata":"https://www.wikidata.org/wiki/Q1161707","display_name":"Benchmark (surveying)","level":2,"score":0.5031999945640564},{"id":"https://openalex.org/C98045186","wikidata":"https://www.wikidata.org/wiki/Q205663","display_name":"Process (computing)","level":2,"score":0.48350000381469727},{"id":"https://openalex.org/C2777212361","wikidata":"https://www.wikidata.org/wiki/Q5127848","display_name":"Class (philosophy)","level":2,"score":0.44929999113082886},{"id":"https://openalex.org/C153180980","wikidata":"https://www.wikidata.org/wiki/Q19776675","display_name":"Commit","level":2,"score":0.41990000009536743},{"id":"https://openalex.org/C45340560","wikidata":"https://www.wikidata.org/wiki/Q215382","display_name":"Disjoint sets","level":2,"score":0.4092000126838684},{"id":"https://openalex.org/C112570922","wikidata":"https://www.wikidata.org/wiki/Q60528603","display_name":"Contamination","level":2,"score":0.3896999955177307},{"id":"https://openalex.org/C106131492","wikidata":"https://www.wikidata.org/wiki/Q3072260","display_name":"Filter (signal processing)","level":2,"score":0.3652999997138977},{"id":"https://openalex.org/C168167062","wikidata":"https://www.wikidata.org/wiki/Q1117970","display_name":"Component (thermodynamics)","level":2,"score":0.361299991607666},{"id":"https://openalex.org/C171268870","wikidata":"https://www.wikidata.org/wiki/Q1486676","display_name":"GRASP","level":2,"score":0.34630000591278076},{"id":"https://openalex.org/C165696696","wikidata":"https://www.wikidata.org/wiki/Q11287","display_name":"Exploit","level":2,"score":0.3199000060558319},{"id":"https://openalex.org/C2779231336","wikidata":"https://www.wikidata.org/wiki/Q7534724","display_name":"Sketch","level":2,"score":0.3197999894618988},{"id":"https://openalex.org/C75291252","wikidata":"https://www.wikidata.org/wiki/Q1315756","display_name":"TRACE (psycholinguistics)","level":2,"score":0.3190000057220459},{"id":"https://openalex.org/C2778826181","wikidata":"https://www.wikidata.org/wiki/Q3310538","display_name":"Culprit","level":3,"score":0.30230000615119934},{"id":"https://openalex.org/C194544171","wikidata":"https://www.wikidata.org/wiki/Q21105679","display_name":"Gating","level":2,"score":0.30090001225471497},{"id":"https://openalex.org/C81917197","wikidata":"https://www.wikidata.org/wiki/Q628760","display_name":"Selection (genetic algorithm)","level":2,"score":0.29919999837875366},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.2890999913215637},{"id":"https://openalex.org/C101468663","wikidata":"https://www.wikidata.org/wiki/Q1620158","display_name":"Modular design","level":2,"score":0.2874999940395355},{"id":"https://openalex.org/C61272859","wikidata":"https://www.wikidata.org/wiki/Q7834031","display_name":"Transferability","level":3,"score":0.2833000123500824},{"id":"https://openalex.org/C31170391","wikidata":"https://www.wikidata.org/wiki/Q188619","display_name":"Hierarchy","level":2,"score":0.28279998898506165},{"id":"https://openalex.org/C153701036","wikidata":"https://www.wikidata.org/wiki/Q659974","display_name":"Trustworthiness","level":2,"score":0.2800999879837036},{"id":"https://openalex.org/C2781238097","wikidata":"https://www.wikidata.org/wiki/Q175026","display_name":"Object (grammar)","level":2,"score":0.2630999982357025},{"id":"https://openalex.org/C124101348","wikidata":"https://www.wikidata.org/wiki/Q172491","display_name":"Data mining","level":1,"score":0.2574000060558319},{"id":"https://openalex.org/C136979486","wikidata":"https://www.wikidata.org/wiki/Q773483","display_name":"Existential quantification","level":2,"score":0.2567000091075897},{"id":"https://openalex.org/C174998907","wikidata":"https://www.wikidata.org/wiki/Q357662","display_name":"Work in process","level":2,"score":0.2554999887943268},{"id":"https://openalex.org/C200601418","wikidata":"https://www.wikidata.org/wiki/Q2193887","display_name":"Reliability engineering","level":1,"score":0.25200000405311584}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:oai:arXiv.org:2608.05810","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2608.05810","pdf_url":"https://arxiv.org/pdf/2608.05810","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":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},{"id":"doi:10.48550/arxiv.2608.05810","doi":"https://doi.org/10.48550/arxiv.2608.05810","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2608.05810","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:2608.05810","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2608.05810","pdf_url":"https://arxiv.org/pdf/2608.05810","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":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},"sustainable_development_goals":[],"study_designs":[],"awards":[],"funders":[],"has_content":{"pdf":true,"grobid_xml":true},"content_urls":{"pdf":"https://content.openalex.org/works/W7201834752.pdf","grobid_xml":"https://content.openalex.org/works/W7201834752.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":["https://openalex.org/W7204167736","https://openalex.org/W7203491292","https://openalex.org/W7164477688","https://openalex.org/W7203492321","https://openalex.org/W7214355539","https://openalex.org/W7164828440","https://openalex.org/W7222614344","https://openalex.org/W7172537726","https://openalex.org/W7167622861","https://openalex.org/W7221150091"],"abstract_inverted_index":{"Self-evolving":[0],"agents":[1],"accumulate":[2],"capability":[3],"by":[4],"distilling":[5,60],"reusable":[6],"skills":[7,27,156,190],"from":[8],"their":[9],"execution":[10],"trajectories,":[11],"but":[12],"we":[13],"find":[14],"this":[15,36],"process":[16],"is":[17,72],"not":[18],"monotonic:":[19],"past":[20],"a":[21,44,48,76,97,108,113,120,143,168,193,215,232,249],"critical":[22],"pool":[23,181,216,224],"size,":[24],"newly":[25],"added":[26],"degrade":[28],"performance":[29],"instead":[30],"of":[31,100,176,187,196,203,253],"improving":[32],"it.":[33],"We":[34,67],"formalize":[35],"capability-contamination":[37],"phase":[38],"transition":[39],"and":[40,116,133,170,184,220,231,244],"trace":[41],"it":[42,55],"to":[43,167,227],"structural":[45,129],"cause:":[46],"once":[47],"defective":[49],"skill":[50,78,106,139,223],"enters":[51],"the":[52,70,80,84,101,152,158,180,188,197,200,239],"decision":[53],"context,":[54],"becomes":[56],"reference":[57],"material":[58],"for":[59],"later":[61],"skills,":[62],"forming":[63],"cross-round":[64],"contamination":[65,71,150],"chains.":[66],"further":[68],"show":[69],"structurally":[73],"irreversible:":[74],"removing":[75],"source":[77],"after":[79],"fact":[81],"cannot":[82],"erase":[83],"flawed":[85],"reasoning":[86],"its":[87,177,221],"descendants":[88],"have":[89],"already":[90],"inherited,":[91],"so":[92],"post-hoc":[93,114,185],"rollback":[94],"recovers":[95,191],"only":[96,192],"small":[98,194],"fraction":[99],"lost":[102],"performance.":[103],"This":[104],"makes":[105],"admission":[107],"pre-commit":[109],"necessity":[110],"rather":[111],"than":[112],"fix,":[115],"motivates":[117],"Verifier-as-Gatekeeper":[118],"(VaG):":[119],"progressive":[121],"trust":[122],"hierarchy":[123],"whose":[124],"three":[125,240],"heterogeneous":[126],"critics":[127,241],"-":[128,136,199],"validity,":[130],"behavioral":[131],"harmlessness,":[132],"semantic":[134],"consistency":[135],"filter":[137],"each":[138,247],"individually,":[140],"coupled":[141],"with":[142,214],"marginal-gain":[144],"subset":[145],"selection":[146],"that":[147],"removes":[148],"combinatorial":[149],"at":[151],"top":[153],"tier":[154],"before":[155],"reach":[157],"runtime":[159],"context.":[160],"On":[161],"Terminal-Bench":[162],"2,":[163],"unconditional":[164],"accumulation":[165],"rises":[166],"peak":[169],"then":[171],"degrades,":[172],"giving":[173],"back":[174],"most":[175],"gains":[178],"as":[179],"keeps":[182],"growing,":[183],"removal":[186],"culprit":[189],"part":[195],"drop":[198],"empirical":[201],"signature":[202],"irreversibility.":[204],"In":[205],"contrast,":[206],"VaG":[207],"improves":[208],"every":[209],"round,":[210],"reaching":[211],"72%":[212],"pass@1":[213],"roughly":[217],"5x":[218],"smaller,":[219],"frozen":[222],"transfers":[225],"positively":[226],"four":[228],"other":[229],"backbones":[230],"second":[233],"benchmark":[234],"without":[235],"re-evolution.":[236],"Ablations":[237],"confirm":[238],"are":[242],"complementary":[243],"mutually":[245],"non-substitutable,":[246],"intercepting":[248],"largely":[250],"disjoint":[251],"class":[252],"harmful":[254],"skills.":[255]},"counts_by_year":[],"updated_date":"2026-10-28T10:33:24.094632","created_date":"2026-08-08T00:00:00"}