{"id":"https://openalex.org/W7204258167","doi":"https://doi.org/10.48550/arxiv.2608.23651","title":"Feedback That Backfires: Why Small Language Model Agents Repeat the Call They Just Watched Fail","display_name":"Feedback That Backfires: Why Small Language Model Agents Repeat the Call They Just Watched Fail","publication_year":2026,"publication_date":"2026-08-24","ids":{"openalex":"https://openalex.org/W7204258167","doi":"https://doi.org/10.48550/arxiv.2608.23651"},"language":null,"primary_location":{"id":"pmh:oai:arXiv.org:2608.23651","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2608.23651","pdf_url":"https://arxiv.org/pdf/2608.23651","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/2608.23651","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5119891973","display_name":"Esmail Gumaan","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I186354981","display_name":"University of Passau","ror":"https://ror.org/05ydjnb78","country_code":"DE","type":"education","lineage":["https://openalex.org/I186354981"]}],"countries":["DE"],"is_corresponding":true,"raw_author_name":"Gumaan, Esmail","raw_affiliation_strings":["Faculty of Computer Science and Mathematics University of Passau , Germany"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Faculty of Computer Science and Mathematics University of Passau , Germany","institution_ids":["https://openalex.org/I186354981"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":1,"corresponding_author_ids":["https://openalex.org/A5119891973"],"corresponding_institution_ids":["https://openalex.org/I186354981"],"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/T10743","display_name":"Software Testing and Debugging Techniques","score":0.37940001487731934,"subfield":{"id":"https://openalex.org/subfields/1712","display_name":"Software"},"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/T10743","display_name":"Software Testing and Debugging Techniques","score":0.37940001487731934,"subfield":{"id":"https://openalex.org/subfields/1712","display_name":"Software"},"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/T10260","display_name":"Software Engineering Research","score":0.20980000495910645,"subfield":{"id":"https://openalex.org/subfields/1710","display_name":"Information Systems"},"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":0.1656000018119812,"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/language-model-agents","display_name":"language model agents","score":0.7509999871253967},{"id":"https://openalex.org/keywords/tool-calling","display_name":"tool calling","score":0.6949999928474426},{"id":"https://openalex.org/keywords/failure-feedback","display_name":"failure feedback","score":0.6230000257492065},{"id":"https://openalex.org/keywords/error-messages","display_name":"error messages","score":0.6359999775886536},{"id":"https://openalex.org/keywords/agent-harnesses","display_name":"agent harnesses","score":0.48399999737739563},{"id":"https://openalex.org/keywords/program-repair","display_name":"program repair","score":0.5989999771118164},{"id":"https://openalex.org/keywords/mbpp","display_name":"MBPP","score":0.8619999885559082},{"id":"https://openalex.org/keywords/greedy-decoding","display_name":"greedy decoding","score":0.4860000014305115}],"concepts":[{"id":"https://openalex.org/C48145219","wikidata":"https://www.wikidata.org/wiki/Q1335365","display_name":"Security token","level":2,"score":0.6333000063896179},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6265000104904175},{"id":"https://openalex.org/C57273362","wikidata":"https://www.wikidata.org/wiki/Q576722","display_name":"Decoding methods","level":2,"score":0.4925000071525574},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.4909000098705292},{"id":"https://openalex.org/C177264268","wikidata":"https://www.wikidata.org/wiki/Q1514741","display_name":"Set (abstract data type)","level":2,"score":0.45329999923706055},{"id":"https://openalex.org/C171268870","wikidata":"https://www.wikidata.org/wiki/Q1486676","display_name":"GRASP","level":2,"score":0.42969998717308044},{"id":"https://openalex.org/C2780791683","wikidata":"https://www.wikidata.org/wiki/Q846785","display_name":"Action (physics)","level":2,"score":0.4194999933242798},{"id":"https://openalex.org/C2777880217","wikidata":"https://www.wikidata.org/wiki/Q4674282","display_name":"Acknowledgement","level":2,"score":0.3765999972820282},{"id":"https://openalex.org/C70437156","wikidata":"https://www.wikidata.org/wiki/Q7228652","display_name":"Pooling","level":2,"score":0.3598000109195709},{"id":"https://openalex.org/C90329073","wikidata":"https://www.wikidata.org/wiki/Q914232","display_name":"Ask price","level":2,"score":0.3598000109195709},{"id":"https://openalex.org/C143095724","wikidata":"https://www.wikidata.org/wiki/Q515895","display_name":"Odds","level":3,"score":0.34200000762939453},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.3246000111103058},{"id":"https://openalex.org/C2776459999","wikidata":"https://www.wikidata.org/wiki/Q2119376","display_name":"Fidelity","level":2,"score":0.31700000166893005},{"id":"https://openalex.org/C184337299","wikidata":"https://www.wikidata.org/wiki/Q1437428","display_name":"Semantics (computer science)","level":2,"score":0.30309998989105225},{"id":"https://openalex.org/C28490314","wikidata":"https://www.wikidata.org/wiki/Q189436","display_name":"Speech recognition","level":1,"score":0.2953000068664551},{"id":"https://openalex.org/C2780009758","wikidata":"https://www.wikidata.org/wiki/Q6804172","display_name":"Measure (data warehouse)","level":2,"score":0.28839999437332153},{"id":"https://openalex.org/C103088060","wikidata":"https://www.wikidata.org/wiki/Q1062839","display_name":"Error detection and correction","level":2,"score":0.272599995136261},{"id":"https://openalex.org/C2777317252","wikidata":"https://www.wikidata.org/wiki/Q18393516","display_name":"Rare events","level":2,"score":0.2694000005722046},{"id":"https://openalex.org/C113954288","wikidata":"https://www.wikidata.org/wiki/Q186885","display_name":"Timestamp","level":2,"score":0.26660001277923584},{"id":"https://openalex.org/C204321447","wikidata":"https://www.wikidata.org/wiki/Q30642","display_name":"Natural language processing","level":1,"score":0.25699999928474426},{"id":"https://openalex.org/C80444323","wikidata":"https://www.wikidata.org/wiki/Q2878974","display_name":"Theoretical computer science","level":1,"score":0.25119999051094055},{"id":"https://openalex.org/C2777200299","wikidata":"https://www.wikidata.org/wiki/Q52943","display_name":"Conversation","level":2,"score":0.25029999017715454},{"id":"https://openalex.org/C2777530160","wikidata":"https://www.wikidata.org/wiki/Q41796","display_name":"Sentence","level":2,"score":0.25029999017715454}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:oai:arXiv.org:2608.23651","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2608.23651","pdf_url":"https://arxiv.org/pdf/2608.23651","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.2608.23651","doi":"https://doi.org/10.48550/arxiv.2608.23651","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2608.23651","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":null,"raw_source_name":null,"raw_type":"Preprint"}],"best_oa_location":{"id":"pmh:oai:arXiv.org:2608.23651","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2608.23651","pdf_url":"https://arxiv.org/pdf/2608.23651","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/W7204258167.pdf","grobid_xml":"https://content.openalex.org/works/W7204258167.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Agent":[0],"harnesses":[1],"record":[2,41],"a":[3,39,94,116,156,159,163,220,274,307],"failed":[4,125,170,187,269],"tool":[5,75],"call":[6,126,154,218],"and":[7,14,77,104,132,190,209,235,266],"its":[8],"error":[9,25,207],"message":[10],"in":[11,45,71,98,192,199],"the":[12,16,21,24,35,43,49,56,85,99,120,124,145,152,169,178,181,200,203,216,224,229,241,245,260,268,277,284,294,298],"transcript":[13],"ask":[15],"model":[17,63],"to":[18,130,271,304],"continue,":[19],"on":[20,106,113,140,244,306],"assumption":[22],"that":[23,51,210,296],"is":[26,58,87,188,198,283],"corrective":[27,36],"information.":[28],"We":[29],"measure":[30],"whether":[31],"it":[32,136,186,264,292],"is.":[33],"Defining":[34],"gain":[37,57],"of":[38,47,96,101,108,122,142,177,184,206,223,228],"failure":[40,146,157,225],"as":[42],"change":[44],"log-probability":[46],"re-emitting":[48],"action":[50,83,92],"just":[52],"failed,":[53],"we":[54,64,287],"find":[55],"negative":[59],"for":[60,138,175,280,289],"every":[61],"instruction-tuned":[62],"tested":[65],"(6":[66],"checkpoints,":[67],"135M-1.7B,":[68],"4":[69],"families)":[70],"two":[72,167],"environments:":[73],"simulated":[74],"calling":[76],"MBPP":[78],"program":[79],"repair.":[80],"Normalised":[81],"by":[82],"length":[84],"effect":[86],"about":[88],"-1.03":[89],"nats":[90],"per":[91],"token,":[93,103],"factor":[95],"2.8":[97],"odds":[100],"each":[102],"holds":[105],"90%-100%":[107],"individual":[109],"items,":[110],"not":[111,202,256],"only":[112],"average.":[114],"Over":[115],"fixed":[117],"candidate":[118],"set":[119],"probability":[121],"repeating":[123],"rises":[127],"from":[128,273],"0.06":[129],"0.54,":[131],"greedy":[133],"decoding":[134],"reproduces":[135],"token":[137,139,233],"19%":[141],"items":[143],"after":[144],"versus":[147],"0%":[148],"before.":[149],"Counterfactuals":[150],"pairing":[151],"same":[153,246],"with":[155,219],"message,":[158,161],"success":[160],"or":[162],"neutral":[164],"acknowledgement":[165],"separate":[166],"effects:":[168],"call's":[171],"surface":[172],"form":[173],"accounts":[174],"83%":[176],"damage,":[179],"while":[180],"semantic":[182],"contribution":[183],"marking":[185],"small":[189],"inconsistent":[191],"sign":[193],"across":[194],"environments.":[195],"The":[196,300],"problem":[197],"harness,":[201],"model's":[204],"grasp":[205],"messages,":[208],"predicts":[211],"which":[212],"remedies":[213,250],"work.":[214],"Replacing":[215],"verbatim":[217],"runtime-generated":[221],"description":[222],"removes":[226],"76%":[227],"inversion":[230],"at":[231,240],"no":[232],"cost,":[234],"making":[236],"previously-failed":[237],"strings":[238],"unreachable":[239],"decoder":[242],"acts":[243],"term.":[247],"Two":[248],"plausible":[249],"do":[251],"not:":[252],"an":[253],"explicit":[254],"\"do":[255],"repeat\"":[257],"instruction":[258],"leaves":[259],"measured":[261,288],"quantity":[262],"where":[263],"was,":[265],"deleting":[267],"attempt":[270],"retry":[272],"clean":[275],"context,":[276],"standard":[278],"prescription":[279],"context":[281,295],"contamination,":[282],"worst":[285],"harness":[286],"repetition,":[290],"because":[291],"restores":[293],"produced":[297],"failure.":[299],"study":[301],"runs":[302],"end":[303,305],"CPU;":[308],"all":[309],"artefacts":[310],"are":[311],"released.":[312]},"counts_by_year":[],"updated_date":"2026-09-30T09:27:27.532063","created_date":"2026-08-27T00:00:00"}