{"id":"https://openalex.org/W7160829750","doi":"https://doi.org/10.48550/arxiv.2605.07177","title":"HyperEyes: Dual-Grained Efficiency-Aware Reinforcement Learning for Parallel Multimodal Search Agents","display_name":"HyperEyes: Dual-Grained Efficiency-Aware Reinforcement Learning for Parallel Multimodal Search Agents","publication_year":2026,"publication_date":"2026-05-08","ids":{"openalex":"https://openalex.org/W7160829750","doi":"https://doi.org/10.48550/arxiv.2605.07177"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2605.07177","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.07177","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":"cc-by","license_id":"https://openalex.org/licenses/cc-by","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.2605.07177","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5135845638","display_name":"Guankai Li","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Li, Guankai","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5112120011","display_name":"Jiabin Chen","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Chen, Jiabin","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5135872081","display_name":"Yi Xu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Xu, Yi","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5135854554","display_name":"Xichen Zhang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhang, Xichen","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5135894246","display_name":"Yuan Lu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Lu, Yuan","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/T11714","display_name":"Multimodal Machine Learning Applications","score":0.8112999796867371,"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"}},"topics":[{"id":"https://openalex.org/T11714","display_name":"Multimodal Machine Learning Applications","score":0.8112999796867371,"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/T10286","display_name":"Information Retrieval and Search Behavior","score":0.05590000003576279,"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/T10028","display_name":"Topic Modeling","score":0.022199999541044235,"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/inference","display_name":"Inference","score":0.6762999892234802},{"id":"https://openalex.org/keywords/reinforcement-learning","display_name":"Reinforcement learning","score":0.6732000112533569},{"id":"https://openalex.org/keywords/benchmark","display_name":"Benchmark (surveying)","score":0.5916000008583069},{"id":"https://openalex.org/keywords/process","display_name":"Process (computing)","score":0.5626000165939331},{"id":"https://openalex.org/keywords/pipeline","display_name":"Pipeline (software)","score":0.5539000034332275},{"id":"https://openalex.org/keywords/trace","display_name":"TRACE (psycholinguistics)","score":0.3693000078201294},{"id":"https://openalex.org/keywords/redundancy","display_name":"Redundancy (engineering)","score":0.34549999237060547}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.8026999831199646},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.6762999892234802},{"id":"https://openalex.org/C97541855","wikidata":"https://www.wikidata.org/wiki/Q830687","display_name":"Reinforcement learning","level":2,"score":0.6732000112533569},{"id":"https://openalex.org/C185798385","wikidata":"https://www.wikidata.org/wiki/Q1161707","display_name":"Benchmark (surveying)","level":2,"score":0.5916000008583069},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.5676000118255615},{"id":"https://openalex.org/C98045186","wikidata":"https://www.wikidata.org/wiki/Q205663","display_name":"Process (computing)","level":2,"score":0.5626000165939331},{"id":"https://openalex.org/C43521106","wikidata":"https://www.wikidata.org/wiki/Q2165493","display_name":"Pipeline (software)","level":2,"score":0.5539000034332275},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.5465999841690063},{"id":"https://openalex.org/C75291252","wikidata":"https://www.wikidata.org/wiki/Q1315756","display_name":"TRACE (psycholinguistics)","level":2,"score":0.3693000078201294},{"id":"https://openalex.org/C152124472","wikidata":"https://www.wikidata.org/wiki/Q1204361","display_name":"Redundancy (engineering)","level":2,"score":0.34549999237060547},{"id":"https://openalex.org/C136197465","wikidata":"https://www.wikidata.org/wiki/Q1729295","display_name":"Variety (cybernetics)","level":2,"score":0.32109999656677246},{"id":"https://openalex.org/C148220186","wikidata":"https://www.wikidata.org/wiki/Q7111912","display_name":"Outcome (game theory)","level":2,"score":0.296999990940094},{"id":"https://openalex.org/C166955791","wikidata":"https://www.wikidata.org/wiki/Q629579","display_name":"Macro","level":2,"score":0.28679999709129333},{"id":"https://openalex.org/C171268870","wikidata":"https://www.wikidata.org/wiki/Q1486676","display_name":"GRASP","level":2,"score":0.28049999475479126},{"id":"https://openalex.org/C22367795","wikidata":"https://www.wikidata.org/wiki/Q7625208","display_name":"Structured prediction","level":2,"score":0.2784999907016754},{"id":"https://openalex.org/C51632099","wikidata":"https://www.wikidata.org/wiki/Q3985153","display_name":"Training set","level":2,"score":0.27720001339912415},{"id":"https://openalex.org/C118505674","wikidata":"https://www.wikidata.org/wiki/Q42586063","display_name":"Encoder","level":2,"score":0.27570000290870667},{"id":"https://openalex.org/C75553542","wikidata":"https://www.wikidata.org/wiki/Q178161","display_name":"A priori and a posteriori","level":2,"score":0.26739999651908875}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2605.07177","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.07177","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":"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":"doi:10.48550/arxiv.2605.07177","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.07177","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":"cc-by","license_id":"https://openalex.org/licenses/cc-by","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":{"Existing":[0],"multimodal":[1,30,54],"search":[2,33,55,70,213],"agents":[3,31],"process":[4],"target":[5],"entities":[6,73],"sequentially,":[7],"issuing":[8],"one":[9],"tool":[10,153],"call":[11],"per":[12],"entity":[13],"and":[14,61,102,215],"accumulating":[15],"redundant":[16],"interaction":[17],"rounds":[18,235],"whenever":[19],"a":[20,44,52,64,79,94,119,140,204],"query":[21],"decomposes":[22],"into":[23,63],"independent":[24],"sub-retrievals.":[25],"We":[26],"argue":[27],"that":[28,57,210],"effective":[29],"should":[32],"wider":[34],"rather":[35],"than":[36],"longer:":[37],"dispatching":[38],"multiple":[39,72],"grounded":[40],"queries":[41],"concurrently":[42],"within":[43],"round.":[45],"To":[46],"this":[47],"end,":[48],"we":[49,92,133,164,201],"present":[50],"HyperEyes,":[51],"parallel":[53],"agent":[56,226],"fuses":[58],"visual":[59,100],"grounding":[60],"retrieval":[62],"single":[65],"atomic":[66],"action,":[67],"enabling":[68],"concurrent":[69],"across":[71],"while":[74],"treating":[75],"inference":[76,199],"efficiency":[77],"as":[78,194],"first-class":[80],"training":[81,149],"objective.":[82],"HyperEyes":[83],"is":[84,145],"trained":[85],"in":[86,229],"two":[87,127],"stages.":[88],"For":[89],"cold-start":[90],"supervision,":[91],"develop":[93],"Parallel-Amenable":[95],"Data":[96],"Synthesis":[97],"Pipeline":[98],"covering":[99],"multi-entity":[101],"textual":[103],"multi-constraint":[104],"queries,":[105],"curating":[106],"efficiency-oriented":[107],"trajectories":[108],"via":[109],"Progressive":[110],"Rejection":[111],"Sampling.":[112],"Building":[113],"on":[114,178,236],"this,":[115],"our":[116],"central":[117],"contribution,":[118],"Dual-Grained":[120],"Efficiency-Aware":[121],"Reinforcement":[122],"Learning":[123],"framework,":[124],"operates":[125],"at":[126],"levels.":[128],"At":[129,160],"the":[130,161,182,195,222],"macro":[131],"level,":[132,163],"propose":[134],"TRACE":[135],"(Tool-use":[136],"Reference-Adaptive":[137],"Cost":[138],"Efficiency),":[139],"trajectory-level":[141],"reward":[142],"whose":[143],"reference":[144],"monotonically":[146],"tightened":[147],"during":[148],"to":[150,168],"suppress":[151],"superfluous":[152],"calls":[154],"without":[155],"restricting":[156],"genuine":[157],"multi-hop":[158],"search.":[159],"micro":[162],"adapt":[165],"On-Policy":[166],"Distillation":[167],"inject":[169],"dense":[170],"token-level":[171],"corrective":[172],"signals":[173],"from":[174],"an":[175],"external":[176],"teacher":[177],"failed":[179],"rollouts,":[180],"mitigating":[181],"credit-assignment":[183],"deficiency":[184],"of":[185,207],"sparse":[186],"outcome":[187],"rewards.":[188],"Since":[189],"existing":[190],"benchmarks":[191],"evaluate":[192],"accuracy":[193,230],"sole":[196],"metric,":[197],"omitting":[198],"cost,":[200],"introduce":[202],"IMEB,":[203],"human-curated":[205],"benchmark":[206],"300":[208],"instances":[209],"jointly":[211],"evaluates":[212],"capability":[214],"efficiency.":[216],"Across":[217],"six":[218],"benchmarks,":[219],"HyperEyes-30B":[220],"surpasses":[221],"strongest":[223],"comparable":[224],"open-source":[225],"by":[227],"9.9%":[228],"with":[231],"5.3x":[232],"fewer":[233],"tool-call":[234],"average.":[237]},"counts_by_year":[],"updated_date":"2026-07-01T06:00:48.157686","created_date":"2026-05-12T00:00:00"}