{"id":"https://openalex.org/W7140141392","doi":"https://doi.org/10.18653/v1/2026.findings-eacl.40","title":"Denoising Concept Vectors with Sparse Autoencoders for Improved Language Model Steering","display_name":"Denoising Concept Vectors with Sparse Autoencoders for Improved Language Model Steering","publication_year":2026,"publication_date":"2026-01-01","ids":{"openalex":"https://openalex.org/W7140141392","doi":"https://doi.org/10.18653/v1/2026.findings-eacl.40"},"language":null,"primary_location":{"id":"doi:10.18653/v1/2026.findings-eacl.40","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2026.findings-eacl.40","pdf_url":"https://aclanthology.org/2026.findings-eacl.40.pdf","source":null,"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Findings of the Association for Computational Linguistics: EACL 2026","raw_type":"proceedings-article"},"type":"conference-paper","indexed_in":["crossref"],"open_access":{"is_oa":true,"oa_status":"gold","oa_url":"https://aclanthology.org/2026.findings-eacl.40.pdf","any_repository_has_fulltext":null},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5130364062","display_name":"Haiyan Zhao","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Haiyan Zhao","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5130393218","display_name":"Xuansheng Wu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Xuansheng Wu","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5130355008","display_name":"Fan Yang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Fan Yang","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5130365401","display_name":"Bo Shen","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Bo Shen","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":null,"display_name":"Ninghao Liu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Ninghao Liu","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5130367629","display_name":"Mengnan Du","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Mengnan Du","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":true,"cited_by_count":0,"citation_normalized_percentile":null,"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"797","last_page":"808"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10664","display_name":"Sentiment Analysis and Opinion Mining","score":0.1046999990940094,"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/T10664","display_name":"Sentiment Analysis and Opinion Mining","score":0.1046999990940094,"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/T10201","display_name":"Speech Recognition and Synthesis","score":0.08540000021457672,"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/T11901","display_name":"Bayesian Methods and Mixture Models","score":0.08179999887943268,"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/noise-reduction","display_name":"Noise reduction","score":0.4896000027656555},{"id":"https://openalex.org/keywords/pattern-recognition","display_name":"Pattern recognition (psychology)","score":0.39340001344680786},{"id":"https://openalex.org/keywords/noise","display_name":"Noise (video)","score":0.37220001220703125},{"id":"https://openalex.org/keywords/language-model","display_name":"Language model","score":0.3671000003814697},{"id":"https://openalex.org/keywords/feature","display_name":"Feature (linguistics)","score":0.3154999911785126}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6503999829292297},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.6294000148773193},{"id":"https://openalex.org/C163294075","wikidata":"https://www.wikidata.org/wiki/Q581861","display_name":"Noise reduction","level":2,"score":0.4896000027656555},{"id":"https://openalex.org/C153180895","wikidata":"https://www.wikidata.org/wiki/Q7148389","display_name":"Pattern recognition (psychology)","level":2,"score":0.39340001344680786},{"id":"https://openalex.org/C99498987","wikidata":"https://www.wikidata.org/wiki/Q2210247","display_name":"Noise (video)","level":3,"score":0.37220001220703125},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.3671000003814697},{"id":"https://openalex.org/C31972630","wikidata":"https://www.wikidata.org/wiki/Q844240","display_name":"Computer vision","level":1,"score":0.34439998865127563},{"id":"https://openalex.org/C2776401178","wikidata":"https://www.wikidata.org/wiki/Q12050496","display_name":"Feature (linguistics)","level":2,"score":0.3154999911785126},{"id":"https://openalex.org/C111335779","wikidata":"https://www.wikidata.org/wiki/Q3454686","display_name":"Reduction (mathematics)","level":2,"score":0.2989000082015991},{"id":"https://openalex.org/C204321447","wikidata":"https://www.wikidata.org/wiki/Q30642","display_name":"Natural language processing","level":1,"score":0.26669999957084656},{"id":"https://openalex.org/C67186912","wikidata":"https://www.wikidata.org/wiki/Q367664","display_name":"Data modeling","level":2,"score":0.26409998536109924},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.25870001316070557},{"id":"https://openalex.org/C28490314","wikidata":"https://www.wikidata.org/wiki/Q189436","display_name":"Speech recognition","level":1,"score":0.25769999623298645}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.18653/v1/2026.findings-eacl.40","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2026.findings-eacl.40","pdf_url":"https://aclanthology.org/2026.findings-eacl.40.pdf","source":null,"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Findings of the Association for Computational Linguistics: EACL 2026","raw_type":"proceedings-article"}],"best_oa_location":{"id":"doi:10.18653/v1/2026.findings-eacl.40","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2026.findings-eacl.40","pdf_url":"https://aclanthology.org/2026.findings-eacl.40.pdf","source":null,"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Findings of the Association for Computational Linguistics: EACL 2026","raw_type":"proceedings-article"},"sustainable_development_goals":[],"awards":[],"funders":[],"has_content":{"grobid_xml":true,"pdf":true},"content_urls":{"pdf":"https://content.openalex.org/works/W7140141392.pdf","grobid_xml":"https://content.openalex.org/works/W7140141392.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Linear":[0],"concept":[1],"vectors":[2],"effectively":[3],"steer":[4],"LLMs,":[5],"but":[6],"existing":[7],"methods":[8],"suffer":[9],"from":[10,48],"noisy":[11],"features":[12],"in":[13],"diverse":[14],"datasets":[15],"that":[16,41,58],"undermine":[17],"steering":[18,73],"robustness.We":[19],"propose":[20],"Sparse":[21],"Autoencoder-Denoised":[22],"Concept":[23],"Vectors":[24],"(SDCV),":[25],"which":[26],"selectively":[27],"keep":[28],"the":[29],"most":[30],"discriminative":[31],"SAE":[32],"latents":[33,57],"while":[34,82],"reconstructing":[35],"hidden":[36],"representations.Our":[37],"key":[38],"insight":[39],"is":[40],"concept-relevant":[42],"signals":[43],"can":[44],"be":[45],"explicitly":[46],"separated":[47],"dataset":[49],"noise":[50],"by":[51,76],"scaling":[52],"up":[53],"activations":[54],"of":[55],"top-k":[56],"best":[59],"differentiate":[60],"positive":[61],"and":[62,68],"negative":[63],"samples.Applied":[64],"to":[65],"linear":[66],"probing":[67],"differencein-mean,":[69],"SDCV":[70],"consistently":[71],"improves":[72],"success":[74],"rates":[75],"4-16%":[77],"across":[78],"six":[79],"challenging":[80],"concepts,":[81],"maintaining":[83],"topic":[84],"relevance.":[85]},"counts_by_year":[],"updated_date":"2026-07-14T23:27:15.235271","created_date":"2026-03-24T00:00:00"}