{"id":"https://openalex.org/W7163043051","doi":"https://doi.org/10.48550/arxiv.2605.31175","title":"Towards Efficient LLMs Annealing with Principled Sample Selection","display_name":"Towards Efficient LLMs Annealing with Principled Sample Selection","publication_year":2026,"publication_date":"2026-05-29","ids":{"openalex":"https://openalex.org/W7163043051","doi":"https://doi.org/10.48550/arxiv.2605.31175"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2605.31175","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.31175","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":null,"license_id":null,"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.2605.31175","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5137516807","display_name":"Yuanjian Xu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Xu, Yuanjian","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5137515672","display_name":"Jianing Hao","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Hao, Jianing","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5137564252","display_name":"Wanbo Zhang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhang, Wanbo","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5137609750","display_name":"Zhong Li","orcid":"https://orcid.org/0000-0001-6087-6854"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Li, Zhong","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5137516640","display_name":"Guang Zhang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhang, Guang","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":5,"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/T11307","display_name":"Domain Adaptation and Few-Shot Learning","score":0.3100000023841858,"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/T11307","display_name":"Domain Adaptation and Few-Shot Learning","score":0.3100000023841858,"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/T10775","display_name":"Generative Adversarial Networks and Image Synthesis","score":0.15369999408721924,"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/T10036","display_name":"Advanced Neural Network Applications","score":0.13130000233650208,"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/simulated-annealing","display_name":"Simulated annealing","score":0.656000018119812},{"id":"https://openalex.org/keywords/gradient-descent","display_name":"Gradient descent","score":0.4925000071525574},{"id":"https://openalex.org/keywords/optimization-problem","display_name":"Optimization problem","score":0.4262000024318695},{"id":"https://openalex.org/keywords/selection","display_name":"Selection (genetic algorithm)","score":0.39959999918937683},{"id":"https://openalex.org/keywords/convergence","display_name":"Convergence (economics)","score":0.38609999418258667},{"id":"https://openalex.org/keywords/sample","display_name":"Sample (material)","score":0.3772999942302704},{"id":"https://openalex.org/keywords/context","display_name":"Context (archaeology)","score":0.3749000132083893}],"concepts":[{"id":"https://openalex.org/C126980161","wikidata":"https://www.wikidata.org/wiki/Q863783","display_name":"Simulated annealing","level":2,"score":0.656000018119812},{"id":"https://openalex.org/C126255220","wikidata":"https://www.wikidata.org/wiki/Q141495","display_name":"Mathematical optimization","level":1,"score":0.5857999920845032},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.574400007724762},{"id":"https://openalex.org/C153258448","wikidata":"https://www.wikidata.org/wiki/Q1199743","display_name":"Gradient descent","level":3,"score":0.4925000071525574},{"id":"https://openalex.org/C137836250","wikidata":"https://www.wikidata.org/wiki/Q984063","display_name":"Optimization problem","level":2,"score":0.4262000024318695},{"id":"https://openalex.org/C81917197","wikidata":"https://www.wikidata.org/wiki/Q628760","display_name":"Selection (genetic algorithm)","level":2,"score":0.39959999918937683},{"id":"https://openalex.org/C2777303404","wikidata":"https://www.wikidata.org/wiki/Q759757","display_name":"Convergence (economics)","level":2,"score":0.38609999418258667},{"id":"https://openalex.org/C198531522","wikidata":"https://www.wikidata.org/wiki/Q485146","display_name":"Sample (material)","level":2,"score":0.3772999942302704},{"id":"https://openalex.org/C2779343474","wikidata":"https://www.wikidata.org/wiki/Q3109175","display_name":"Context (archaeology)","level":2,"score":0.3749000132083893},{"id":"https://openalex.org/C55660270","wikidata":"https://www.wikidata.org/wiki/Q5164377","display_name":"Constrained optimization","level":2,"score":0.36730000376701355},{"id":"https://openalex.org/C93959086","wikidata":"https://www.wikidata.org/wiki/Q6888345","display_name":"Model selection","level":2,"score":0.2854999899864197},{"id":"https://openalex.org/C164752517","wikidata":"https://www.wikidata.org/wiki/Q5570875","display_name":"Global optimization","level":2,"score":0.28299999237060547},{"id":"https://openalex.org/C46714192","wikidata":"https://www.wikidata.org/wiki/Q4680763","display_name":"Adaptive simulated annealing","level":3,"score":0.27230000495910645},{"id":"https://openalex.org/C26517878","wikidata":"https://www.wikidata.org/wiki/Q228039","display_name":"Key (lock)","level":2,"score":0.26750001311302185},{"id":"https://openalex.org/C206688291","wikidata":"https://www.wikidata.org/wiki/Q7617819","display_name":"Stochastic gradient descent","level":3,"score":0.26420000195503235},{"id":"https://openalex.org/C2989514635","wikidata":"https://www.wikidata.org/wiki/Q5164377","display_name":"Constrained optimization problem","level":3,"score":0.2556999921798706}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2605.31175","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.31175","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":null,"license_id":null,"version":null,"is_accepted":false,"is_published":null,"raw_source_name":null,"raw_type":"article"}],"best_oa_location":{"id":"doi:10.48550/arxiv.2605.31175","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.31175","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":null,"license_id":null,"version":null,"is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"article"},"sustainable_development_goals":[],"awards":[],"funders":[],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"The":[0],"annealing":[1,56,116],"phase":[2,24,57],"is":[3,167],"a":[4,26,44,91,107,119],"pivotal":[5],"convergence":[6,71],"stage":[7,117],"in":[8,47,114],"LLM":[9],"pre-training":[10],"that":[11,69,110,142,157],"ultimately":[12],"determines":[13],"final":[14],"model":[15,154],"quality.":[16],"However,":[17],"effectively":[18],"selecting":[19],"training":[20],"data":[21,88],"during":[22],"this":[23,51,84,99],"remains":[25],"key":[27],"challenge.":[28],"Current":[29],"strategies":[30],"rely":[31],"on":[32,83,128,132],"empirical":[33],"heuristics,":[34],"such":[35],"as":[36,90,118],"domain":[37],"filtering":[38],"or":[39],"context":[40],"extension,":[41],"which":[42],"lack":[43],"principled":[45],"grounding":[46],"optimization":[48,121],"theory.":[49],"In":[50],"work,":[52],"we":[53,86,101],"characterize":[54],"the":[55,59,62,115,133,137,145],"through":[58],"lens":[60],"of":[61,93,136],"loss":[63],"landscape's":[64],"spectral":[65,134],"geometry.":[66],"We":[67],"argue":[68],"optimal":[70,146],"requires":[72],"gradient":[73],"updates":[74],"to":[75],"satisfy":[76],"heterogeneous":[77],"constraints":[78,127],"across":[79,152],"different":[80],"eigen-directions.":[81],"Building":[82],"insight,":[85],"formulate":[87],"selection":[89,113],"problem":[92],"satisfying":[94],"these":[95],"directional":[96,126],"constraints.":[97],"To":[98],"end,":[100],"propose":[102],"DiReCT":[103,139,158],"(Directionally-Restrained":[104],"Constrained":[105],"Training),":[106],"novel":[108],"framework":[109],"reformulates":[111],"sample":[112],"constrained":[120],"problem.":[122],"By":[123],"imposing":[124],"explicit":[125],"per-sample":[129],"gradients":[130],"based":[131],"properties":[135],"Hessian,":[138],"identifies":[140],"samples":[141],"align":[143],"with":[144],"curvature-aware":[147],"descent":[148],"path.":[149],"Extensive":[150],"experiments":[151],"various":[153],"scales":[155],"demonstrate":[156],"consistently":[159],"achieves":[160],"state-of-the-art":[161],"performance.":[162],"For":[163],"future":[164],"research,":[165],"code":[166],"available":[168],"at":[169],"https://github.com/xuyj233/Direct.":[170]},"counts_by_year":[],"updated_date":"2026-06-11T09:08:48.828518","created_date":"2026-06-02T00:00:00"}