{"id":"https://openalex.org/W7128775868","doi":"https://doi.org/10.48550/arxiv.2602.11767","title":"TSR: Trajectory-Search Rollouts for Multi-Turn RL of LLM Agents","display_name":"TSR: Trajectory-Search Rollouts for Multi-Turn RL of LLM Agents","publication_year":2026,"publication_date":"2026-02-12","ids":{"openalex":"https://openalex.org/W7128775868","doi":"https://doi.org/10.48550/arxiv.2602.11767"},"language":"en","primary_location":{"id":"pmh:oai:arXiv.org:2602.11767","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2602.11767","pdf_url":"https://arxiv.org/pdf/2602.11767","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/2602.11767","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5094040082","display_name":"Aladin Djuhera","orcid":"https://orcid.org/0009-0005-1641-8801","observed_orcids":["https://orcid.org/0009-0005-1641-8801"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Djuhera, Aladin","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5053049612","display_name":"Swanand Kadhe","orcid":"https://orcid.org/0000-0003-2006-1030","observed_orcids":["https://orcid.org/0000-0003-2006-1030"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Kadhe, Swanand","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5122974693","display_name":"Farhan Ahmed","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Ahmed, Farhan","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5119875612","display_name":"Syed Zawad","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zawad, Syed","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5048671912","display_name":"Heiko Ludwig","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Ludwig, Heiko","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5146835314","display_name":"Holger Boche","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Boche, Holger","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/T10462","display_name":"Reinforcement Learning in Robotics","score":0.9994999766349792,"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/T10462","display_name":"Reinforcement Learning in Robotics","score":0.9994999766349792,"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/T11574","display_name":"Artificial Intelligence in Games","score":9.999999747378752e-05,"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":9.999999747378752e-05,"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/multi-turn-reinforcement-learning","display_name":"multi-turn reinforcement learning","score":0.906000018119812},{"id":"https://openalex.org/keywords/llm-based-agents","display_name":"LLM agents","score":0.8510000109672546},{"id":"https://openalex.org/keywords/trajectory-search","display_name":"trajectory search","score":0.7599999904632568},{"id":"https://openalex.org/keywords/tree-search","display_name":"tree search","score":0.652999997138977},{"id":"https://openalex.org/keywords/beam-search","display_name":"beam search","score":0.8009999990463257},{"id":"https://openalex.org/keywords/sparse-rewards","display_name":"sparse rewards","score":0.6660000085830688},{"id":"https://openalex.org/keywords/mode-collapse","display_name":"mode collapse","score":0.7310000061988831},{"id":"https://openalex.org/keywords/ppo","display_name":"PPO","score":0.6129999756813049},{"id":"https://openalex.org/keywords/group-relative-policy-optimization","display_name":"Group Relative Policy Optimization","score":0.9100000262260437},{"id":"https://openalex.org/keywords/sokoban","display_name":"Sokoban","score":0.6830000281333923},{"id":"https://openalex.org/keywords/webshop","display_name":"WebShop","score":0.8119999766349792}],"concepts":[{"id":"https://openalex.org/C97541855","wikidata":"https://www.wikidata.org/wiki/Q830687","display_name":"Reinforcement learning","level":2,"score":0.8004000186920166},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7353000044822693},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.5906999707221985},{"id":"https://openalex.org/C2780801425","wikidata":"https://www.wikidata.org/wiki/Q5164392","display_name":"Construct (python library)","level":2,"score":0.5552999973297119},{"id":"https://openalex.org/C13662910","wikidata":"https://www.wikidata.org/wiki/Q193139","display_name":"Trajectory","level":2,"score":0.5551999807357788},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.5498999953269958},{"id":"https://openalex.org/C81917197","wikidata":"https://www.wikidata.org/wiki/Q628760","display_name":"Selection (genetic algorithm)","level":2,"score":0.5109000205993652},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.45660001039505005},{"id":"https://openalex.org/C2779530757","wikidata":"https://www.wikidata.org/wiki/Q1207505","display_name":"Quality (philosophy)","level":2,"score":0.4147000014781952},{"id":"https://openalex.org/C192209626","wikidata":"https://www.wikidata.org/wiki/Q190909","display_name":"Focus (optics)","level":2,"score":0.40630000829696655},{"id":"https://openalex.org/C89611455","wikidata":"https://www.wikidata.org/wiki/Q6804646","display_name":"Mechanism (biology)","level":2,"score":0.36910000443458557},{"id":"https://openalex.org/C140779682","wikidata":"https://www.wikidata.org/wiki/Q210868","display_name":"Sampling (signal processing)","level":3,"score":0.3659999966621399},{"id":"https://openalex.org/C2778755073","wikidata":"https://www.wikidata.org/wiki/Q10858537","display_name":"Scale (ratio)","level":2,"score":0.3580000102519989},{"id":"https://openalex.org/C99844830","wikidata":"https://www.wikidata.org/wiki/Q102441924","display_name":"Scaling","level":2,"score":0.3237000107765198},{"id":"https://openalex.org/C2780586882","wikidata":"https://www.wikidata.org/wiki/Q7520643","display_name":"Simple (philosophy)","level":2,"score":0.31859999895095825},{"id":"https://openalex.org/C26517878","wikidata":"https://www.wikidata.org/wiki/Q228039","display_name":"Key (lock)","level":2,"score":0.273499995470047},{"id":"https://openalex.org/C111335779","wikidata":"https://www.wikidata.org/wiki/Q3454686","display_name":"Reduction (mathematics)","level":2,"score":0.2680000066757202}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:oai:arXiv.org:2602.11767","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2602.11767","pdf_url":"https://arxiv.org/pdf/2602.11767","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.2602.11767","doi":"https://doi.org/10.48550/arxiv.2602.11767","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2602.11767","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:2602.11767","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2602.11767","pdf_url":"https://arxiv.org/pdf/2602.11767","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":{"grobid_xml":true,"pdf":true},"content_urls":{"pdf":"https://content.openalex.org/works/W7128775868.pdf","grobid_xml":"https://content.openalex.org/works/W7128775868.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":["https://openalex.org/W7169806788","https://openalex.org/W4416146915","https://openalex.org/W4417167303","https://openalex.org/W4401103013","https://openalex.org/W7139926955","https://openalex.org/W7102410087","https://openalex.org/W4414971194","https://openalex.org/W4414810684","https://openalex.org/W7167709631","https://openalex.org/W7203904183"],"abstract_inverted_index":{"Advances":[0],"in":[1,126],"large":[2],"language":[3],"models":[4],"(LLMs)":[5],"are":[6,31],"driving":[7],"a":[8,59,160],"shift":[9],"toward":[10],"using":[11],"reinforcement":[12],"learning":[13],"(RL)":[14],"to":[15,77,120,152,169],"train":[16],"agents":[17],"from":[18,149],"iterative,":[19],"multi-turn":[20,25,165],"interactions":[21],"across":[22],"tasks.":[23],"However,":[24],"RL":[26],"remains":[27,101],"challenging":[28],"as":[29],"rewards":[30],"often":[32],"sparse":[33],"or":[34],"delayed,":[35],"and":[36,50,85,100,112,124,172],"environments":[37],"can":[38,47],"be":[39],"stochastic.":[40],"In":[41],"this":[42],"regime,":[43],"naive":[44],"trajectory":[45,86],"sampling":[46],"hinder":[48],"exploitation":[49],"induce":[51],"mode":[52],"collapse.":[53],"We":[54],"propose":[55],"TSR":[56,72,114,158],"(Trajectory-Search":[57],"Rollouts),":[58],"training-time":[60,133],"approach":[61],"that":[62,139],"repurposes":[63],"test-time":[64],"scaling":[65],"ideas":[66],"for":[67,136,163],"improved":[68],"per-turn":[69],"rollout":[70,89,93,134,154],"generation.":[71,90],"performs":[73],"lightweight":[74],"tree-style":[75],"search":[76,142,148],"construct":[78],"higher-quality":[79],"trajectories":[80],"by":[81,107],"selecting":[82],"promising":[83],"actions":[84],"prefixes":[87],"during":[88],"This":[91],"improves":[92],"quality":[94],"while":[95,130],"preserving":[96],"stable":[97],"policy":[98],"optimization":[99,128],"compatible":[102],"with":[103],"standard":[104],"policy-gradient":[105],"optimizers":[106],"design.":[108],"Across":[109],"Sokoban,":[110],"FrozenLake,":[111],"WebShop,":[113],"achieves":[115],"success-rate":[116],"gains":[117],"of":[118,156],"up":[119],"15":[121],"percentage":[122],"points":[123],"converges":[125],"fewer":[127],"steps,":[129],"trading":[131],"additional":[132],"compute":[135],"stronger":[137,164],"policies":[138],"require":[140],"no":[141],"at":[143],"inference":[144],"time.":[145],"By":[146],"moving":[147],"test":[150],"time":[151],"the":[153],"stage":[155],"training,":[157],"provides":[159],"modular":[161],"mechanism":[162],"agent":[166],"learning,":[167],"complementary":[168],"existing":[170],"frameworks":[171],"rejection-sampling-style":[173],"selection":[174],"methods.":[175]},"counts_by_year":[],"updated_date":"2026-10-28T10:33:24.094632","created_date":"2026-02-14T00:00:00"}