{"id":"https://openalex.org/W7133318532","doi":"https://doi.org/10.48550/arxiv.2603.01416","title":"Securing the Floor and Raising the Ceiling: A Merging-based Paradigm for Multi-modal Search Agents","display_name":"Securing the Floor and Raising the Ceiling: A Merging-based Paradigm for Multi-modal Search Agents","publication_year":2026,"publication_date":"2026-03-02","ids":{"openalex":"https://openalex.org/W7133318532","doi":"https://doi.org/10.48550/arxiv.2603.01416"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2603.01416","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2603.01416","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":"article"},"type":"preprint","indexed_in":["datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://doi.org/10.48550/arxiv.2603.01416","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5127933573","display_name":"Zhixiang Wang","orcid":null},"institutions":[],"countries":[],"is_corresponding":true,"raw_author_name":"Wang, Zhixiang","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5128019912","display_name":"Jingxuan Xu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Xu, Jingxuan","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5103155183","display_name":"Dajun Chen","orcid":"https://orcid.org/0009-0000-9532-7636"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Chen, Dajun","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5127914386","display_name":"Yunfang Wu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wu, Yunfang","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5127981618","display_name":"Wei Jiang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Jiang, Wei","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5127873298","display_name":"Yong Li","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Li, Yong","raw_affiliation_strings":[],"affiliations":[]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":6,"corresponding_author_ids":["https://openalex.org/A5127933573"],"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":0.9394999742507935,"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":0.9394999742507935,"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/T10028","display_name":"Topic Modeling","score":0.008899999782443047,"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/T10627","display_name":"Advanced Image and Video Retrieval Techniques","score":0.004900000058114529,"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"}}],"keywords":[{"id":"https://openalex.org/keywords/set","display_name":"Set (abstract data type)","score":0.5903000235557556},{"id":"https://openalex.org/keywords/reinforcement-learning","display_name":"Reinforcement learning","score":0.5231000185012817},{"id":"https://openalex.org/keywords/convergence","display_name":"Convergence (economics)","score":0.4925999939441681},{"id":"https://openalex.org/keywords/ceiling","display_name":"Ceiling (cloud)","score":0.4027000069618225},{"id":"https://openalex.org/keywords/search-algorithm","display_name":"Search algorithm","score":0.3619000017642975},{"id":"https://openalex.org/keywords/base","display_name":"Base (topology)","score":0.35420000553131104},{"id":"https://openalex.org/keywords/raising","display_name":"Raising (metalworking)","score":0.33390000462532043}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6266999840736389},{"id":"https://openalex.org/C177264268","wikidata":"https://www.wikidata.org/wiki/Q1514741","display_name":"Set (abstract data type)","level":2,"score":0.5903000235557556},{"id":"https://openalex.org/C97541855","wikidata":"https://www.wikidata.org/wiki/Q830687","display_name":"Reinforcement learning","level":2,"score":0.5231000185012817},{"id":"https://openalex.org/C2777303404","wikidata":"https://www.wikidata.org/wiki/Q759757","display_name":"Convergence (economics)","level":2,"score":0.4925999939441681},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.4456000030040741},{"id":"https://openalex.org/C2777489069","wikidata":"https://www.wikidata.org/wiki/Q1589822","display_name":"Ceiling (cloud)","level":2,"score":0.4027000069618225},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.39320001006126404},{"id":"https://openalex.org/C125583679","wikidata":"https://www.wikidata.org/wiki/Q755673","display_name":"Search algorithm","level":2,"score":0.3619000017642975},{"id":"https://openalex.org/C42058472","wikidata":"https://www.wikidata.org/wiki/Q810214","display_name":"Base (topology)","level":2,"score":0.35420000553131104},{"id":"https://openalex.org/C2780589192","wikidata":"https://www.wikidata.org/wiki/Q7285140","display_name":"Raising (metalworking)","level":2,"score":0.33390000462532043},{"id":"https://openalex.org/C2777211547","wikidata":"https://www.wikidata.org/wiki/Q17141490","display_name":"Training (meteorology)","level":2,"score":0.33340001106262207},{"id":"https://openalex.org/C26517878","wikidata":"https://www.wikidata.org/wiki/Q228039","display_name":"Key (lock)","level":2,"score":0.3255999982357025},{"id":"https://openalex.org/C51632099","wikidata":"https://www.wikidata.org/wiki/Q3985153","display_name":"Training set","level":2,"score":0.3093000054359436},{"id":"https://openalex.org/C201789804","wikidata":"https://www.wikidata.org/wiki/Q2362762","display_name":"Search problem","level":2,"score":0.28450000286102295},{"id":"https://openalex.org/C126661757","wikidata":"https://www.wikidata.org/wiki/Q4925641","display_name":"Random search","level":2,"score":0.2770000100135803},{"id":"https://openalex.org/C165838908","wikidata":"https://www.wikidata.org/wiki/Q736777","display_name":"Calibration","level":2,"score":0.2623000144958496}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2603.01416","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2603.01416","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":"article"}],"best_oa_location":{"id":"doi:10.48550/arxiv.2603.01416","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2603.01416","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":"article"},"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":{"Recent":[0],"advances":[1],"in":[2],"Vision-Language":[3],"Models":[4],"(VLMs)":[5],"have":[6],"motivated":[7],"the":[8,167],"development":[9],"of":[10,132],"multi-modal":[11,86,96],"search":[12,19,67,77,87,161],"agents":[13],"that":[14,85,116],"can":[15,89],"actively":[16],"invoke":[17],"external":[18],"tools":[20],"and":[21,49,177],"integrate":[22],"retrieved":[23],"evidence":[24],"through":[25],"multi-step":[26],"reasoning.":[27],"While":[28],"promising,":[29],"existing":[30],"approaches":[31],"typically":[32],"rely":[33],"on":[34,121,124,137],"large-scale":[35],"supervised":[36],"trajectories":[37],"or":[38],"expensive":[39],"reinforcement":[40],"learning":[41],"(RL),":[42],"leading":[43],"to":[44,62],"high":[45],"training":[46,97],"cost,":[47],"instability,":[48],"a":[50,59,75,80,112,129,149,154,171],"severe":[51],"cold-start":[52],"problem":[53],"for":[54],"standard":[55,182],"VLMs.":[56],"We":[57],"propose":[58],"training-free":[60],"paradigm":[61],"empower":[63],"VLMs":[64],"with":[65,79,157],"autonomous":[66],"capabilities":[68,88],"via":[69],"cross-modal":[70,104],"model":[71,125],"merging.":[72],"By":[73],"fusing":[74],"text-based":[76],"agent":[78],"base":[81],"VLM,":[82],"we":[83,106],"show":[84],"be":[90],"effectively":[91],"composed":[92],"without":[93],"any":[94],"additional":[95],"data.":[98],"To":[99],"mitigate":[100],"parameter":[101],"interference":[102],"during":[103],"integration,":[105],"introduce":[107],"Optimal":[108],"Brain":[109],"Merging":[110],"(OBM),":[111],"saliency-aware":[113],"merging":[114,147],"algorithm":[115],"identifies":[117],"task-critical":[118],"parameters":[119],"based":[120],"their":[122],"impact":[123],"loss":[126],"using":[127],"only":[128],"small":[130],"set":[131],"calibration":[133],"samples.":[134],"Extensive":[135],"experiments":[136],"search-intensive":[138],"benchmarks":[139],"(e.g.,":[140],"InfoSeek,":[141],"MMSearch)":[142],"reveal":[143],"that:":[144],"(1)":[145],"Model":[146],"secures":[148],"reasonable":[150],"performance":[151,168],"floor":[152],"as":[153,170],"zero-shot":[155],"agent,":[156],"OBM":[158,164],"achieving":[159,174],"superior":[160],"rates;":[162],"(2)":[163],"significantly":[165],"raises":[166],"ceiling":[169],"warm-start":[172],"strategy,":[173],"faster":[175],"convergence":[176],"higher":[178],"peak":[179],"accuracy":[180],"than":[181],"VLM":[183],"initialization.":[184]},"counts_by_year":[],"updated_date":"2026-03-04T07:09:34.246503","created_date":"2026-03-04T00:00:00"}