{"id":"https://openalex.org/W7131832694","doi":"https://doi.org/10.48550/arxiv.2602.22871","title":"Test-Time Scaling with Diffusion Language Models via Reward-Guided Stitching","display_name":"Test-Time Scaling with Diffusion Language Models via Reward-Guided Stitching","publication_year":2026,"publication_date":"2026-02-26","ids":{"openalex":"https://openalex.org/W7131832694","doi":"https://doi.org/10.48550/arxiv.2602.22871"},"language":null,"primary_location":{"id":"pmh:doi:10.48550/arxiv.2602.22871","is_oa":true,"landing_page_url":null,"pdf_url":null,"source":{"id":"https://openalex.org/S4406922384","display_name":"Open MIND","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"host_organization_lineage_names":[],"type":"repository"},"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"Article"},"type":"article","indexed_in":["datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":null,"any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5127052861","display_name":"Roy Miles","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Miles, Roy","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5051910500","display_name":"Aysim Toker","orcid":"https://orcid.org/0000-0002-5391-9403"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Toker, Aysim","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5127120939","display_name":"Andreea-Maria Oncescu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Oncescu, Andreea-Maria","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5126985605","display_name":"Songcen Xu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Xu, Songcen","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5127358444","display_name":"Jiankang Deng","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Deng, Jiankang","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5127465731","display_name":"Ismail Elezi","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Elezi, Ismail","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":0.0,"has_fulltext":false,"cited_by_count":0,"citation_normalized_percentile":{"value":0.24383649,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"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/T10028","display_name":"Topic Modeling","score":0.5659000277519226,"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/T10028","display_name":"Topic Modeling","score":0.5659000277519226,"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/T11714","display_name":"Multimodal Machine Learning Applications","score":0.11330000311136246,"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/T11273","display_name":"Advanced Graph Neural Networks","score":0.04859999939799309,"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/modular-design","display_name":"Modular design","score":0.5347999930381775},{"id":"https://openalex.org/keywords/language-model","display_name":"Language model","score":0.49000000953674316},{"id":"https://openalex.org/keywords/image-stitching","display_name":"Image stitching","score":0.4652000069618225},{"id":"https://openalex.org/keywords/solver","display_name":"Solver","score":0.4415999948978424},{"id":"https://openalex.org/keywords/pipeline","display_name":"Pipeline (software)","score":0.41780000925064087},{"id":"https://openalex.org/keywords/scaling","display_name":"Scaling","score":0.39809998869895935},{"id":"https://openalex.org/keywords/coding","display_name":"Coding (social sciences)","score":0.3659999966621399},{"id":"https://openalex.org/keywords/encoder","display_name":"Encoder","score":0.3628999888896942},{"id":"https://openalex.org/keywords/imperfect","display_name":"Imperfect","score":0.3571999967098236},{"id":"https://openalex.org/keywords/scalability","display_name":"Scalability","score":0.34790000319480896}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7024999856948853},{"id":"https://openalex.org/C101468663","wikidata":"https://www.wikidata.org/wiki/Q1620158","display_name":"Modular design","level":2,"score":0.5347999930381775},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.49000000953674316},{"id":"https://openalex.org/C29081049","wikidata":"https://www.wikidata.org/wiki/Q1364242","display_name":"Image stitching","level":2,"score":0.4652000069618225},{"id":"https://openalex.org/C2778770139","wikidata":"https://www.wikidata.org/wiki/Q1966904","display_name":"Solver","level":2,"score":0.4415999948978424},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.4278999865055084},{"id":"https://openalex.org/C43521106","wikidata":"https://www.wikidata.org/wiki/Q2165493","display_name":"Pipeline (software)","level":2,"score":0.41780000925064087},{"id":"https://openalex.org/C99844830","wikidata":"https://www.wikidata.org/wiki/Q102441924","display_name":"Scaling","level":2,"score":0.39809998869895935},{"id":"https://openalex.org/C179518139","wikidata":"https://www.wikidata.org/wiki/Q5140297","display_name":"Coding (social sciences)","level":2,"score":0.3659999966621399},{"id":"https://openalex.org/C118505674","wikidata":"https://www.wikidata.org/wiki/Q42586063","display_name":"Encoder","level":2,"score":0.3628999888896942},{"id":"https://openalex.org/C2780310539","wikidata":"https://www.wikidata.org/wiki/Q12547192","display_name":"Imperfect","level":2,"score":0.3571999967098236},{"id":"https://openalex.org/C80444323","wikidata":"https://www.wikidata.org/wiki/Q2878974","display_name":"Theoretical computer science","level":1,"score":0.35089999437332153},{"id":"https://openalex.org/C48044578","wikidata":"https://www.wikidata.org/wiki/Q727490","display_name":"Scalability","level":2,"score":0.34790000319480896},{"id":"https://openalex.org/C140779682","wikidata":"https://www.wikidata.org/wiki/Q210868","display_name":"Sampling (signal processing)","level":3,"score":0.3269999921321869},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.31940001249313354},{"id":"https://openalex.org/C137800194","wikidata":"https://www.wikidata.org/wiki/Q11713455","display_name":"Interpolation (computer graphics)","level":3,"score":0.31690001487731934},{"id":"https://openalex.org/C69357855","wikidata":"https://www.wikidata.org/wiki/Q163214","display_name":"Diffusion","level":2,"score":0.31029999256134033},{"id":"https://openalex.org/C520049643","wikidata":"https://www.wikidata.org/wiki/Q189760","display_name":"Voting","level":3,"score":0.30820000171661377},{"id":"https://openalex.org/C2780069185","wikidata":"https://www.wikidata.org/wiki/Q7977945","display_name":"Equivalence (formal languages)","level":2,"score":0.3034000098705292},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.2989000082015991},{"id":"https://openalex.org/C98045186","wikidata":"https://www.wikidata.org/wiki/Q205663","display_name":"Process (computing)","level":2,"score":0.29429998993873596},{"id":"https://openalex.org/C159877910","wikidata":"https://www.wikidata.org/wiki/Q2202883","display_name":"Autoregressive model","level":2,"score":0.29100000858306885},{"id":"https://openalex.org/C111335779","wikidata":"https://www.wikidata.org/wiki/Q3454686","display_name":"Reduction (mathematics)","level":2,"score":0.28279998898506165},{"id":"https://openalex.org/C125411270","wikidata":"https://www.wikidata.org/wiki/Q18653","display_name":"Encoding (memory)","level":2,"score":0.2784000039100647},{"id":"https://openalex.org/C174348530","wikidata":"https://www.wikidata.org/wiki/Q188635","display_name":"Bridging (networking)","level":2,"score":0.27730000019073486},{"id":"https://openalex.org/C43126263","wikidata":"https://www.wikidata.org/wiki/Q128751","display_name":"Source code","level":2,"score":0.2703000009059906},{"id":"https://openalex.org/C75291252","wikidata":"https://www.wikidata.org/wiki/Q1315756","display_name":"TRACE (psycholinguistics)","level":2,"score":0.26969999074935913},{"id":"https://openalex.org/C149810388","wikidata":"https://www.wikidata.org/wiki/Q5374873","display_name":"Emulation","level":2,"score":0.2685999870300293},{"id":"https://openalex.org/C98763669","wikidata":"https://www.wikidata.org/wiki/Q176645","display_name":"Markov chain","level":2,"score":0.2614000141620636},{"id":"https://openalex.org/C52740198","wikidata":"https://www.wikidata.org/wiki/Q1539564","display_name":"Importance sampling","level":3,"score":0.25679999589920044},{"id":"https://openalex.org/C57273362","wikidata":"https://www.wikidata.org/wiki/Q576722","display_name":"Decoding methods","level":2,"score":0.2531999945640564}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:doi:10.48550/arxiv.2602.22871","is_oa":true,"landing_page_url":null,"pdf_url":null,"source":{"id":"https://openalex.org/S4406922384","display_name":"Open MIND","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"host_organization_lineage_names":[],"type":"repository"},"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"Article"},{"id":"doi:10.48550/arxiv.2602.22871","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2602.22871","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":"pmh:doi:10.48550/arxiv.2602.22871","is_oa":true,"landing_page_url":null,"pdf_url":null,"source":{"id":"https://openalex.org/S4406922384","display_name":"Open MIND","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"host_organization_lineage_names":[],"type":"repository"},"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"Article"},"sustainable_development_goals":[{"display_name":"Quality Education","score":0.4621778726577759,"id":"https://metadata.un.org/sdg/4"}],"awards":[],"funders":[],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Reasoning":[0],"with":[1,82,173],"large":[2],"language":[3,75],"models":[4,210],"often":[5],"benefits":[6],"from":[7,33,122],"generating":[8],"multiple":[9],"chains-of-thought,":[10],"but":[11,163],"existing":[12],"aggregation":[13],"strategies":[14],"are":[15],"typically":[16],"trajectory-level":[17],"(e.g.,":[18,211,217],"selecting":[19],"the":[20,26,113,153,156,194],"best":[21],"trace":[22],"or":[23,35],"voting":[24],"on":[25,147],"final":[27,114,157],"answer),":[28],"discarding":[29],"useful":[30],"intermediate":[31,80],"work":[32],"partial":[34],"\"nearly":[36],"correct\"":[37],"attempts.":[38],"We":[39],"propose":[40],"Stitching":[41],"Noisy":[42],"Diffusion":[43],"Thoughts,":[44],"a":[45,54,61,72,98,201],"self-consistency":[46],"framework":[47,179],"that":[48,141],"turns":[49],"cheap":[50],"diffusion-sampled":[51],"reasoning":[52,69,137],"into":[53,97,166],"reusable":[55],"pool":[56],"of":[57,155],"step-level":[58,142],"candidates.":[59],"Given":[60],"problem,":[62],"we":[63,139],"(i)":[64],"sample":[65],"many":[66],"diverse,":[67],"low-cost":[68],"trajectories":[70,96],"using":[71],"masked":[73],"diffusion":[74,171,209],"model,":[76],"(ii)":[77],"score":[78],"every":[79],"step":[81],"an":[83,105],"off-the-shelf":[84],"process":[85],"reward":[86],"model":[87,108],"(PRM),":[88],"and":[89,124,150,190,214],"(iii)":[90],"stitch":[91],"these":[92],"highest-quality":[93],"steps":[94],"across":[95,187],"composite":[99],"rationale.":[100],"This":[101,116],"rationale":[102],"then":[103],"conditions":[104],"autoregressive":[106],"(AR)":[107],"(solver)":[109],"to":[110,185,200,206],"recompute":[111],"only":[112],"answer.":[115],"modular":[117],"pipeline":[118],"separates":[119],"exploration":[120],"(diffusion)":[121],"evaluation":[123],"solution":[125],"synthesis,":[126],"avoiding":[127],"monolithic":[128],"unified":[129,215],"hybrids":[130],"while":[131],"preserving":[132],"broad":[133],"search.":[134],"Across":[135],"math":[136,189],"benchmarks,":[138],"find":[140],"recombination":[143],"is":[144,220],"most":[145],"beneficial":[146],"harder":[148],"problems,":[149],"ablations":[151],"highlight":[152],"importance":[154],"AR":[158],"solver":[159],"in":[160],"converting":[161],"stitched":[162],"imperfect":[164],"rationales":[165],"accurate":[167],"answers.":[168],"Using":[169],"low-confidence":[170],"sampling":[172],"parallel,":[174],"independent":[175],"rollouts,":[176],"our":[177],"training-free":[178],"improves":[180],"average":[181],"accuracy":[182],"by":[183],"up":[184,199],"23.8%":[186],"six":[188],"coding":[191],"tasks.":[192],"At":[193],"same":[195],"time,":[196],"it":[197],"achieves":[198],"1.8x":[202],"latency":[203],"reduction":[204],"relative":[205],"both":[207],"traditional":[208],"Dream,":[212],"LLaDA)":[213],"architectures":[216],"TiDAR).":[218],"Code":[219],"available":[221],"at":[222],"https://github.com/roymiles/diffusion-stitching.":[223]},"counts_by_year":[],"updated_date":"2026-07-15T18:14:33.161393","created_date":"2026-02-28T00:00:00"}