{"id":"https://openalex.org/W7169777848","doi":"https://doi.org/10.48550/arxiv.2607.15555","title":"LLMs Encode Relevance as a Layer-Wise Cross-Lingual Signal","display_name":"LLMs Encode Relevance as a Layer-Wise Cross-Lingual Signal","publication_year":2026,"publication_date":"2026-07-17","ids":{"openalex":"https://openalex.org/W7169777848","doi":"https://doi.org/10.48550/arxiv.2607.15555"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2607.15555","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2607.15555","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,"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.2607.15555","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5114601870","display_name":"Pietro Bernardelle","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Bernardelle, Pietro","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5120727123","display_name":"Samaneh Mohtadi","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Mohtadi, Samaneh","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5115563658","display_name":"Stefano Civelli","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Civelli, Stefano","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5140780162","display_name":"Joel Mackenzie","orcid":"https://orcid.org/0000-0001-7992-4633"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Mackenzie, Joel","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5141211595","display_name":"Gianluca Demartini","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Demartini, Gianluca","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/T10286","display_name":"Information Retrieval and Search Behavior","score":0.4993000030517578,"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"}},"topics":[{"id":"https://openalex.org/T10286","display_name":"Information Retrieval and Search Behavior","score":0.4993000030517578,"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":0.156700000166893,"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/T13629","display_name":"Text Readability and Simplification","score":0.08940000087022781,"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/relevance","display_name":"Relevance (law)","score":0.9373000264167786},{"id":"https://openalex.org/keywords/encode","display_name":"ENCODE","score":0.5202999711036682},{"id":"https://openalex.org/keywords/relevance-theory","display_name":"Relevance theory","score":0.4277999997138977},{"id":"https://openalex.org/keywords/perspective","display_name":"Perspective (graphical)","score":0.4253000020980835},{"id":"https://openalex.org/keywords/representation","display_name":"Representation (politics)","score":0.3163999915122986}],"concepts":[{"id":"https://openalex.org/C158154518","wikidata":"https://www.wikidata.org/wiki/Q7310970","display_name":"Relevance (law)","level":2,"score":0.9373000264167786},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.527999997138977},{"id":"https://openalex.org/C66746571","wikidata":"https://www.wikidata.org/wiki/Q1134833","display_name":"ENCODE","level":3,"score":0.5202999711036682},{"id":"https://openalex.org/C2776802673","wikidata":"https://www.wikidata.org/wiki/Q1548396","display_name":"Relevance theory","level":3,"score":0.4277999997138977},{"id":"https://openalex.org/C12713177","wikidata":"https://www.wikidata.org/wiki/Q1900281","display_name":"Perspective (graphical)","level":2,"score":0.4253000020980835},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.3903000056743622},{"id":"https://openalex.org/C180747234","wikidata":"https://www.wikidata.org/wiki/Q23373","display_name":"Cognitive psychology","level":1,"score":0.38280001282691956},{"id":"https://openalex.org/C204321447","wikidata":"https://www.wikidata.org/wiki/Q30642","display_name":"Natural language processing","level":1,"score":0.373199999332428},{"id":"https://openalex.org/C15744967","wikidata":"https://www.wikidata.org/wiki/Q9418","display_name":"Psychology","level":0,"score":0.34200000762939453},{"id":"https://openalex.org/C2776359362","wikidata":"https://www.wikidata.org/wiki/Q2145286","display_name":"Representation (politics)","level":3,"score":0.3163999915122986},{"id":"https://openalex.org/C2779843651","wikidata":"https://www.wikidata.org/wiki/Q7390335","display_name":"SIGNAL (programming language)","level":2,"score":0.3125},{"id":"https://openalex.org/C23123220","wikidata":"https://www.wikidata.org/wiki/Q816826","display_name":"Information retrieval","level":1,"score":0.275299996137619},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.2667999863624573},{"id":"https://openalex.org/C63363279","wikidata":"https://www.wikidata.org/wiki/Q5133848","display_name":"Clinical significance","level":2,"score":0.2662999927997589}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2607.15555","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2607.15555","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,"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.2607.15555","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2607.15555","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,"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":[{"display_name":"Quality Education","score":0.6615537405014038,"id":"https://metadata.un.org/sdg/4"}],"awards":[],"funders":[],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Large":[0],"language":[1],"models":[2],"(LLMs)":[3],"are":[4],"increasingly":[5],"used":[6],"in":[7,54,143,148,163],"information":[8],"retrieval":[9],"(IR)":[10],"pipelines":[11],"as":[12,135],"relevance":[13,33,47,63,87,103,132,153,183,210,218,231],"judges":[14],"and":[15,65,75,97,112,146,173,185,229,237],"re-rankers.":[16],"Yet":[17],"most":[18],"analyses":[19],"remain":[20],"output-centric,":[21],"evaluating":[22],"generated":[23,62,110,171,221],"labels":[24],"or":[25,169],"scores":[26],"while":[27],"offering":[28],"limited":[29],"insight":[30],"into":[31],"how":[32,57,230],"is":[34,48,133,141],"represented":[35],"inside":[36],"the":[37,72],"model.":[38],"In":[39],"this":[40,58,202],"work,":[41],"we":[42,79],"study":[43],"whether":[44,66,118],"query-document":[45],"(q-d)":[46],"linearly":[49,156],"decodable":[50],"from":[51,93],"residual-stream":[52],"activations":[53,92],"instruction-tuned":[55],"LLMs,":[56],"signal":[59],"compares":[60],"with":[61,85,109],"judgments,":[64],"it":[67],"transfers":[68],"across":[69,234],"languages.":[70],"Using":[71],"TREC":[73,114],"DL20":[74,115],"MIRACL":[76],"evaluation":[77],"collections,":[78],"guide":[80],"medium-scale":[81],"LLMs":[82],"(4-9B":[83],"parameters)":[84],"UMBRELA-style":[86],"judgment":[88],"prompts,":[89],"extract":[90],"last-token":[91],"every":[94],"transformer":[95],"layer,":[96],"train":[98],"linear":[99],"probes":[100,167],"to":[101,116,224],"predict":[102],"labels.":[104],"We":[105],"compare":[106],"probe":[107,139],"predictions":[108],"judgments":[111,172,222],"use":[113],"test":[117],"probe-derived":[119],"pseudo-labels":[120],"preserve":[121,175],"system":[122,176],"rankings":[123],"against":[124],"human":[125],"judgments.":[126],"Our":[127],"results":[128],"suggest":[129,190],"that":[130,152],"q-d":[131],"encoded":[134],"a":[136,179,205],"depth-dependent":[137],"signal:":[138],"performance":[140],"weak":[142],"early":[144],"layers":[145],"strongest":[147],"middle-to-late":[149],"layers,":[150],"indicating":[151],"becomes":[154],"more":[155],"accessible":[157],"after":[158],"contextual":[159],"integration.":[160],"Most":[161],"importantly,":[162],"several":[164],"models,":[165],"validation-selected":[166],"match":[168],"outperform":[170],"better":[174],"rankings,":[177],"revealing":[178],"separation":[180],"between":[181],"internal":[182],"representation":[184],"external":[186],"expression.":[187],"Multilingual":[188],"experiments":[189],"partial":[191],"cross-language":[192],"portability,":[193],"although":[194],"transfer":[195],"remains":[196],"weaker":[197],"than":[198],"within-language":[199],"decoding.":[200],"Overall,":[201],"work":[203],"provides":[204],"representation-level":[206],"perspective":[207],"on":[208],"LLM-based":[209],"assessment.":[211],"Layer-wise":[212],"probing":[213],"can":[214],"help":[215],"diagnose":[216],"where":[217],"emerges,":[219],"when":[220],"fail":[223],"reflect":[225],"internally":[226],"available":[227],"evidence,":[228],"representations":[232],"vary":[233],"languages,":[235],"datasets,":[236],"model":[238],"families.":[239]},"counts_by_year":[],"updated_date":"2026-07-28T07:46:37.118299","created_date":"2026-07-21T00:00:00"}