{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:TDXHM335BVPTIKPEO7BZHACZEZ","short_pith_number":"pith:TDXHM335","schema_version":"1.0","canonical_sha256":"98ee766f7d0d5f3429e477c393805926532e685bb94c37f760a1876508868947","source":{"kind":"arxiv","id":"2401.13544","version":3},"attestation_state":"computed","paper":{"title":"Beyond Concept Bottleneck Models: How to Make Black Boxes Intervenable?","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["stat.ML"],"primary_cat":"cs.LG","authors_text":"Julia E. Vogt, Moritz Vandenhirtz, Ri\\v{c}ards Marcinkevi\\v{c}s, Sonia Laguna","submitted_at":"2024-01-24T16:02:14Z","abstract_excerpt":"Recently, interpretable machine learning has re-explored concept bottleneck models (CBM). An advantage of this model class is the user's ability to intervene on predicted concept values, affecting the downstream output. In this work, we introduce a method to perform such concept-based interventions on pretrained neural networks, which are not interpretable by design, only given a small validation set with concept labels. Furthermore, we formalise the notion of intervenability as a measure of the effectiveness of concept-based interventions and leverage this definition to fine-tune black boxes."},"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":"2401.13544","kind":"arxiv","version":3},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2024-01-24T16:02:14Z","cross_cats_sorted":["stat.ML"],"title_canon_sha256":"35326c22a3c388ee8bb652a7c3b73079e271cc5190631f2872e588ca7771d498","abstract_canon_sha256":"45fcb1ee3c6a225af98e946a3c34e97e00d01353c5eb9b0807a7866eedb04a0c"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T09:26:23.630823Z","signature_b64":"pmPS49eTUp1bEMzXEkdD5zggz1V66nLu4G6UsYo8l/rB4OdyBVFkiR7mHqMxoyZrcJWNoRKOjm0KpobjueHLAA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"98ee766f7d0d5f3429e477c393805926532e685bb94c37f760a1876508868947","last_reissued_at":"2026-07-05T09:26:23.630404Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T09:26:23.630404Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Beyond Concept Bottleneck Models: How to Make Black Boxes Intervenable?","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["stat.ML"],"primary_cat":"cs.LG","authors_text":"Julia E. Vogt, Moritz Vandenhirtz, Ri\\v{c}ards Marcinkevi\\v{c}s, Sonia Laguna","submitted_at":"2024-01-24T16:02:14Z","abstract_excerpt":"Recently, interpretable machine learning has re-explored concept bottleneck models (CBM). An advantage of this model class is the user's ability to intervene on predicted concept values, affecting the downstream output. In this work, we introduce a method to perform such concept-based interventions on pretrained neural networks, which are not interpretable by design, only given a small validation set with concept labels. Furthermore, we formalise the notion of intervenability as a measure of the effectiveness of concept-based interventions and leverage this definition to fine-tune black boxes."},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2401.13544","kind":"arxiv","version":3},"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/2401.13544/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":"2401.13544","created_at":"2026-07-05T09:26:23.630462+00:00"},{"alias_kind":"arxiv_version","alias_value":"2401.13544v3","created_at":"2026-07-05T09:26:23.630462+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2401.13544","created_at":"2026-07-05T09:26:23.630462+00:00"},{"alias_kind":"pith_short_12","alias_value":"TDXHM335BVPT","created_at":"2026-07-05T09:26:23.630462+00:00"},{"alias_kind":"pith_short_16","alias_value":"TDXHM335BVPTIKPE","created_at":"2026-07-05T09:26:23.630462+00:00"},{"alias_kind":"pith_short_8","alias_value":"TDXHM335","created_at":"2026-07-05T09:26:23.630462+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2411.19395","citing_title":"Concept-driven Off Policy Evaluation","ref_index":2014,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/TDXHM335BVPTIKPEO7BZHACZEZ","json":"https://pith.science/pith/TDXHM335BVPTIKPEO7BZHACZEZ.json","graph_json":"https://pith.science/api/pith-number/TDXHM335BVPTIKPEO7BZHACZEZ/graph.json","events_json":"https://pith.science/api/pith-number/TDXHM335BVPTIKPEO7BZHACZEZ/events.json","paper":"https://pith.science/paper/TDXHM335"},"agent_actions":{"view_html":"https://pith.science/pith/TDXHM335BVPTIKPEO7BZHACZEZ","download_json":"https://pith.science/pith/TDXHM335BVPTIKPEO7BZHACZEZ.json","view_paper":"https://pith.science/paper/TDXHM335","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2401.13544&json=true","fetch_graph":"https://pith.science/api/pith-number/TDXHM335BVPTIKPEO7BZHACZEZ/graph.json","fetch_events":"https://pith.science/api/pith-number/TDXHM335BVPTIKPEO7BZHACZEZ/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/TDXHM335BVPTIKPEO7BZHACZEZ/action/timestamp_anchor","attest_storage":"https://pith.science/pith/TDXHM335BVPTIKPEO7BZHACZEZ/action/storage_attestation","attest_author":"https://pith.science/pith/TDXHM335BVPTIKPEO7BZHACZEZ/action/author_attestation","sign_citation":"https://pith.science/pith/TDXHM335BVPTIKPEO7BZHACZEZ/action/citation_signature","submit_replication":"https://pith.science/pith/TDXHM335BVPTIKPEO7BZHACZEZ/action/replication_record"}},"created_at":"2026-07-05T09:26:23.630462+00:00","updated_at":"2026-07-05T09:26:23.630462+00:00"}