{"id":"https://openalex.org/W7161984852","doi":"https://doi.org/10.48550/arxiv.2605.20745","title":"The Hidden Signal of Verifier Strictness: Controlling and Improving Step-Wise Verification via Selective Latent Steering","display_name":"The Hidden Signal of Verifier Strictness: Controlling and Improving Step-Wise Verification via Selective Latent Steering","publication_year":2026,"publication_date":"2026-05-20","ids":{"openalex":"https://openalex.org/W7161984852","doi":"https://doi.org/10.48550/arxiv.2605.20745"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2605.20745","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.20745","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.2605.20745","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5136701316","display_name":"Yefan Zhou","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhou, Yefan","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5074749353","display_name":"Yilun Zhou","orcid":"https://orcid.org/0000-0002-7085-3880"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhou, Yilun","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5002266393","display_name":"Austin Xu","orcid":"https://orcid.org/0009-0003-1694-1841"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Xu, Austin","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5035399743","display_name":"Soroush Vosoughi","orcid":"https://orcid.org/0000-0002-2564-8909"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Vosoughi, Soroush","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5136717637","display_name":"Shafiq Joty","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Joty, Shafiq","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5136725452","display_name":"Jiang Gui","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Gui, Jiang","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/T10142","display_name":"Formal Methods in Verification","score":0.21250000596046448,"subfield":{"id":"https://openalex.org/subfields/1703","display_name":"Computational Theory and Mathematics"},"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/T10142","display_name":"Formal Methods in Verification","score":0.21250000596046448,"subfield":{"id":"https://openalex.org/subfields/1703","display_name":"Computational Theory and Mathematics"},"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/T11689","display_name":"Adversarial Robustness in Machine Learning","score":0.16169999539852142,"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/T11424","display_name":"Security and Verification in Computing","score":0.12120000272989273,"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/correctness","display_name":"Correctness","score":0.8830000162124634},{"id":"https://openalex.org/keywords/inference","display_name":"Inference","score":0.574400007724762},{"id":"https://openalex.org/keywords/code","display_name":"Code (set theory)","score":0.4763000011444092},{"id":"https://openalex.org/keywords/exploit","display_name":"Exploit","score":0.45559999346733093},{"id":"https://openalex.org/keywords/signal","display_name":"SIGNAL (programming language)","score":0.453000009059906},{"id":"https://openalex.org/keywords/encoding","display_name":"Encoding (memory)","score":0.39899998903274536},{"id":"https://openalex.org/keywords/key","display_name":"Key (lock)","score":0.37940001487731934}],"concepts":[{"id":"https://openalex.org/C55439883","wikidata":"https://www.wikidata.org/wiki/Q360812","display_name":"Correctness","level":2,"score":0.8830000162124634},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.8324999809265137},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.574400007724762},{"id":"https://openalex.org/C2776760102","wikidata":"https://www.wikidata.org/wiki/Q5139990","display_name":"Code (set theory)","level":3,"score":0.4763000011444092},{"id":"https://openalex.org/C165696696","wikidata":"https://www.wikidata.org/wiki/Q11287","display_name":"Exploit","level":2,"score":0.45559999346733093},{"id":"https://openalex.org/C2779843651","wikidata":"https://www.wikidata.org/wiki/Q7390335","display_name":"SIGNAL (programming language)","level":2,"score":0.453000009059906},{"id":"https://openalex.org/C125411270","wikidata":"https://www.wikidata.org/wiki/Q18653","display_name":"Encoding (memory)","level":2,"score":0.39899998903274536},{"id":"https://openalex.org/C26517878","wikidata":"https://www.wikidata.org/wiki/Q228039","display_name":"Key (lock)","level":2,"score":0.37940001487731934},{"id":"https://openalex.org/C80444323","wikidata":"https://www.wikidata.org/wiki/Q2878974","display_name":"Theoretical computer science","level":1,"score":0.3772999942302704},{"id":"https://openalex.org/C66746571","wikidata":"https://www.wikidata.org/wiki/Q1134833","display_name":"ENCODE","level":3,"score":0.3734999895095825},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.36410000920295715},{"id":"https://openalex.org/C2775924081","wikidata":"https://www.wikidata.org/wiki/Q55608371","display_name":"Control (management)","level":2,"score":0.32199999690055847},{"id":"https://openalex.org/C120314980","wikidata":"https://www.wikidata.org/wiki/Q180634","display_name":"Distributed computing","level":1,"score":0.30079999566078186},{"id":"https://openalex.org/C167966045","wikidata":"https://www.wikidata.org/wiki/Q5532625","display_name":"Generative model","level":3,"score":0.29580000042915344},{"id":"https://openalex.org/C126042441","wikidata":"https://www.wikidata.org/wiki/Q1324888","display_name":"Frame (networking)","level":2,"score":0.2770000100135803},{"id":"https://openalex.org/C202973057","wikidata":"https://www.wikidata.org/wiki/Q7380130","display_name":"Runtime verification","level":3,"score":0.2639000117778778},{"id":"https://openalex.org/C137836250","wikidata":"https://www.wikidata.org/wiki/Q984063","display_name":"Optimization problem","level":2,"score":0.26100000739097595},{"id":"https://openalex.org/C62460635","wikidata":"https://www.wikidata.org/wiki/Q5508853","display_name":"Functional verification","level":3,"score":0.25040000677108765}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2605.20745","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.20745","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.2605.20745","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.20745","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":[{"id":"https://metadata.un.org/sdg/16","display_name":"Peace, Justice and strong institutions","score":0.5908642411231995}],"awards":[],"funders":[],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Generative":[0],"verifiers":[1],"have":[2],"emerged":[3],"as":[4,45],"a":[5,64,71,78,110],"promising":[6],"paradigm":[7],"for":[8,129],"step-wise":[9,69],"verification,":[10,70],"but":[11],"their":[12],"verification":[13,89,169],"behavior":[14],"is":[15,81,154,165,181],"often":[16],"poorly":[17],"calibrated:":[18],"they":[19],"may":[20],"be":[21,39,57],"under-critical":[22],"and":[23,29,115,132,141,149,153],"miss":[24],"erroneous":[25],"steps,":[26],"or":[27,42,76],"over-critical":[28],"reject":[30,77],"correct":[31],"reasoning.":[32],"We":[33,62],"refer":[34],"to":[35,38,74,168],"this":[36,49,92],"tendency":[37,73],"overly":[40,43],"lenient":[41],"critical":[44],"verifier":[46,54,102],"strictness.":[47],"In":[48],"work,":[50],"we":[51,94,121],"study":[52],"whether":[53],"strictness":[55,103],"can":[56,99],"controlled":[58],"through":[59],"hidden-state":[60,66,97],"intervention.":[61],"uncover":[63],"verification-specific":[65],"signal:":[67],"in":[68],"verifier's":[72],"accept":[75],"solution":[79],"step":[80],"encoded":[82],"near":[83],"the":[84,87],"boundary":[85],"of":[86,176],"corresponding":[88],"paragraph.":[90],"Exploiting":[91],"signal,":[93],"show":[95,143],"that":[96,144],"steering":[98,108,151],"directly":[100],"modulate":[101],"without":[104],"fine-tuning.":[105],"However,":[106],"uniform":[107],"induces":[109],"trade-off":[111],"between":[112],"error":[113],"detection":[114],"correctness":[116,127],"certification.":[117],"To":[118],"address":[119],"this,":[120],"propose":[122],"VerifySteer,":[123],"which":[124],"exploits":[125],"latent":[126],"signals":[128],"sample-level":[130],"routing":[131],"selectively":[133],"intervenes":[134],"on":[135,139,174],"paragraph":[136],"boundaries.":[137],"Experiments":[138],"ProcessBench":[140],"Hard2Verify":[142],"VerifySteer":[145,164],"outperforms":[146],"prompt":[147],"optimization":[148],"activation":[150],"baselines,":[152],"competitive":[155],"with":[156],"self-consistency":[157],"while":[158],"requiring":[159],"4-7x":[160],"less":[161],"inference":[162],"compute.":[163],"also":[166],"complementary":[167],"fine-tuning,":[170],"providing":[171],"further":[172],"gains":[173],"top":[175],"fine-tuned":[177],"verifiers.":[178],"The":[179],"code":[180],"available":[182],"at":[183],"https://github.com/YefanZhou/VerifySteer.":[184]},"counts_by_year":[],"updated_date":"2026-07-01T06:00:48.157686","created_date":"2026-05-22T00:00:00"}