{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:IRCKLKJBD7T7V77LSUFNRTT3ER","short_pith_number":"pith:IRCKLKJB","schema_version":"1.0","canonical_sha256":"4444a5a9211fe7faffeb950ad8ce7b245cabd5413e4af7db8ca9d47d9bd1a6ce","source":{"kind":"arxiv","id":"2605.23645","version":1},"attestation_state":"computed","paper":{"title":"Learning Through Noise: Why Subliminal Learning Works and When It Fails","license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Bel\\'en Hidalgo-Ogalde, Roman D. Ventzke, Valentin Neuhaus, Vincent C. Brockers, Viola Priesemann","submitted_at":"2026-05-22T13:59:13Z","abstract_excerpt":"In the context of artificial neural networks, subliminal learning refers to the transfer of task-relevant knowledge or unintended biases from teacher to student models through distillation on task-unrelated input$\\unicode{x2013}$output pairs. Prior explanations tie this effect to shared or closely matched teacher$\\unicode{x2013}$student initialization. We show that a closely matched initialization is not necessary. Instead, subliminal learning is governed by compatible output heads. Using a controlled MNIST setting, we split outputs into an auxiliary head (for auxiliary, task-unrelated noise s"},"verification_status":{"content_addressed":true,"pith_receipt":true,"author_attested":false,"weak_author_claims":0,"strong_author_claims":0,"externally_anchored":false,"storage_verified":false,"citation_signatures":0,"replication_records":0,"graph_snapshot":true,"references_resolved":false,"formal_links_present":false},"canonical_record":{"source":{"id":"2605.23645","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","primary_cat":"cs.LG","submitted_at":"2026-05-22T13:59:13Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"1eb1d4dd2e60e5eedcf9be548b187ef0587023c5eb86fc6cce8c0fbdc336309d","abstract_canon_sha256":"67e56c38063c49940770ef7072abda0e27d2887df66bbacd3febe215068aea8b"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-25T02:02:23.652499Z","signature_b64":"pzRoQI5PD+gthLSWhEmij2yDQf9lvnO4UspTxVrBCTkTRA+FbHP8IqWAbmwdlX4o6VRD4VgJdy1VUB7e5JbdDw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"4444a5a9211fe7faffeb950ad8ce7b245cabd5413e4af7db8ca9d47d9bd1a6ce","last_reissued_at":"2026-05-25T02:02:23.651781Z","signature_status":"signed_v1","first_computed_at":"2026-05-25T02:02:23.651781Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Learning Through Noise: Why Subliminal Learning Works and When It Fails","license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Bel\\'en Hidalgo-Ogalde, Roman D. Ventzke, Valentin Neuhaus, Vincent C. Brockers, Viola Priesemann","submitted_at":"2026-05-22T13:59:13Z","abstract_excerpt":"In the context of artificial neural networks, subliminal learning refers to the transfer of task-relevant knowledge or unintended biases from teacher to student models through distillation on task-unrelated input$\\unicode{x2013}$output pairs. Prior explanations tie this effect to shared or closely matched teacher$\\unicode{x2013}$student initialization. We show that a closely matched initialization is not necessary. Instead, subliminal learning is governed by compatible output heads. Using a controlled MNIST setting, we split outputs into an auxiliary head (for auxiliary, task-unrelated noise s"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2605.23645","kind":"arxiv","version":1},"verdict":{"id":null,"model_set":{},"created_at":null,"strongest_claim":"","one_line_summary":"","pipeline_version":null,"weakest_assumption":"","pith_extraction_headline":""},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2605.23645/integrity.json","findings":[],"available":true,"detectors_run":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938"},"references":{"count":0,"sample":[],"resolved_work":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","internal_anchors":0},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"},"aliases":[{"alias_kind":"arxiv","alias_value":"2605.23645","created_at":"2026-05-25T02:02:23.651879+00:00"},{"alias_kind":"arxiv_version","alias_value":"2605.23645v1","created_at":"2026-05-25T02:02:23.651879+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.23645","created_at":"2026-05-25T02:02:23.651879+00:00"},{"alias_kind":"pith_short_12","alias_value":"IRCKLKJBD7T7","created_at":"2026-05-25T02:02:23.651879+00:00"},{"alias_kind":"pith_short_16","alias_value":"IRCKLKJBD7T7V77L","created_at":"2026-05-25T02:02:23.651879+00:00"},{"alias_kind":"pith_short_8","alias_value":"IRCKLKJB","created_at":"2026-05-25T02:02:23.651879+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/IRCKLKJBD7T7V77LSUFNRTT3ER","json":"https://pith.science/pith/IRCKLKJBD7T7V77LSUFNRTT3ER.json","graph_json":"https://pith.science/api/pith-number/IRCKLKJBD7T7V77LSUFNRTT3ER/graph.json","events_json":"https://pith.science/api/pith-number/IRCKLKJBD7T7V77LSUFNRTT3ER/events.json","paper":"https://pith.science/paper/IRCKLKJB"},"agent_actions":{"view_html":"https://pith.science/pith/IRCKLKJBD7T7V77LSUFNRTT3ER","download_json":"https://pith.science/pith/IRCKLKJBD7T7V77LSUFNRTT3ER.json","view_paper":"https://pith.science/paper/IRCKLKJB","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2605.23645&json=true","fetch_graph":"https://pith.science/api/pith-number/IRCKLKJBD7T7V77LSUFNRTT3ER/graph.json","fetch_events":"https://pith.science/api/pith-number/IRCKLKJBD7T7V77LSUFNRTT3ER/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/IRCKLKJBD7T7V77LSUFNRTT3ER/action/timestamp_anchor","attest_storage":"https://pith.science/pith/IRCKLKJBD7T7V77LSUFNRTT3ER/action/storage_attestation","attest_author":"https://pith.science/pith/IRCKLKJBD7T7V77LSUFNRTT3ER/action/author_attestation","sign_citation":"https://pith.science/pith/IRCKLKJBD7T7V77LSUFNRTT3ER/action/citation_signature","submit_replication":"https://pith.science/pith/IRCKLKJBD7T7V77LSUFNRTT3ER/action/replication_record"}},"created_at":"2026-05-25T02:02:23.651879+00:00","updated_at":"2026-05-25T02:02:23.651879+00:00"}