{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:6CW4QNHYPEO66ISYLPYMNOYXYC","short_pith_number":"pith:6CW4QNHY","schema_version":"1.0","canonical_sha256":"f0adc834f8791def22585bf0c6bb17c08809c6a89aa76e01ed7edfc327cafc79","source":{"kind":"arxiv","id":"2407.00886","version":3},"attestation_state":"computed","paper":{"title":"Efficient Automated Circuit Discovery in Transformers using Contextual Decomposition","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CL","cs.LG"],"primary_cat":"cs.AI","authors_text":"Aliyah R. Hsu, Anobel Y. Odisho, Bin Yu, Georgia Zhou, Peter R. Carroll, Yaxuan Huang, Yeshwanth Cherapanamjeri","submitted_at":"2024-07-01T01:12:20Z","abstract_excerpt":"Automated mechanistic interpretation research has attracted great interest due to its potential to scale explanations of neural network internals to large models. Existing automated circuit discovery work relies on activation patching or its approximations to identify subgraphs in models for specific tasks (circuits). They often suffer from slow runtime, approximation errors, and specific requirements of metrics, such as non-zero gradients. In this work, we introduce contextual decomposition for transformers (CD-T) to build interpretable circuits in large language models. CD-T can produce circ"},"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":"2407.00886","kind":"arxiv","version":3},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.AI","submitted_at":"2024-07-01T01:12:20Z","cross_cats_sorted":["cs.CL","cs.LG"],"title_canon_sha256":"a1d6b776cbe7e39aa53751b797766451e9871f891a5d7f1f879323285f885ba5","abstract_canon_sha256":"9d333200d1b1b14b757da91ba5e880e18dcd55df0902084d3d64e87155164b9b"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:22:01.747754Z","signature_b64":"jPYu5agLufeG2C/ZWWSzKoCtDyQ0CWmCI/gSFBobrVPy+pRVjHzrX5HPYHFE69ghfXKcVHIYRjb5G1shfmzZBw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"f0adc834f8791def22585bf0c6bb17c08809c6a89aa76e01ed7edfc327cafc79","last_reissued_at":"2026-07-05T10:22:01.747202Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:22:01.747202Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Efficient Automated Circuit Discovery in Transformers using Contextual Decomposition","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CL","cs.LG"],"primary_cat":"cs.AI","authors_text":"Aliyah R. Hsu, Anobel Y. Odisho, Bin Yu, Georgia Zhou, Peter R. Carroll, Yaxuan Huang, Yeshwanth Cherapanamjeri","submitted_at":"2024-07-01T01:12:20Z","abstract_excerpt":"Automated mechanistic interpretation research has attracted great interest due to its potential to scale explanations of neural network internals to large models. Existing automated circuit discovery work relies on activation patching or its approximations to identify subgraphs in models for specific tasks (circuits). They often suffer from slow runtime, approximation errors, and specific requirements of metrics, such as non-zero gradients. In this work, we introduce contextual decomposition for transformers (CD-T) to build interpretable circuits in large language models. CD-T can produce circ"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2407.00886","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/2407.00886/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":"2407.00886","created_at":"2026-07-05T10:22:01.747265+00:00"},{"alias_kind":"arxiv_version","alias_value":"2407.00886v3","created_at":"2026-07-05T10:22:01.747265+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2407.00886","created_at":"2026-07-05T10:22:01.747265+00:00"},{"alias_kind":"pith_short_12","alias_value":"6CW4QNHYPEO6","created_at":"2026-07-05T10:22:01.747265+00:00"},{"alias_kind":"pith_short_16","alias_value":"6CW4QNHYPEO66ISY","created_at":"2026-07-05T10:22:01.747265+00:00"},{"alias_kind":"pith_short_8","alias_value":"6CW4QNHY","created_at":"2026-07-05T10:22:01.747265+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2607.07316","citing_title":"Mechanistic Interpretability for Neural Networks: Circuits, Sparse Features and Symbolic Reasoning","ref_index":22,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/6CW4QNHYPEO66ISYLPYMNOYXYC","json":"https://pith.science/pith/6CW4QNHYPEO66ISYLPYMNOYXYC.json","graph_json":"https://pith.science/api/pith-number/6CW4QNHYPEO66ISYLPYMNOYXYC/graph.json","events_json":"https://pith.science/api/pith-number/6CW4QNHYPEO66ISYLPYMNOYXYC/events.json","paper":"https://pith.science/paper/6CW4QNHY"},"agent_actions":{"view_html":"https://pith.science/pith/6CW4QNHYPEO66ISYLPYMNOYXYC","download_json":"https://pith.science/pith/6CW4QNHYPEO66ISYLPYMNOYXYC.json","view_paper":"https://pith.science/paper/6CW4QNHY","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2407.00886&json=true","fetch_graph":"https://pith.science/api/pith-number/6CW4QNHYPEO66ISYLPYMNOYXYC/graph.json","fetch_events":"https://pith.science/api/pith-number/6CW4QNHYPEO66ISYLPYMNOYXYC/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/6CW4QNHYPEO66ISYLPYMNOYXYC/action/timestamp_anchor","attest_storage":"https://pith.science/pith/6CW4QNHYPEO66ISYLPYMNOYXYC/action/storage_attestation","attest_author":"https://pith.science/pith/6CW4QNHYPEO66ISYLPYMNOYXYC/action/author_attestation","sign_citation":"https://pith.science/pith/6CW4QNHYPEO66ISYLPYMNOYXYC/action/citation_signature","submit_replication":"https://pith.science/pith/6CW4QNHYPEO66ISYLPYMNOYXYC/action/replication_record"}},"created_at":"2026-07-05T10:22:01.747265+00:00","updated_at":"2026-07-05T10:22:01.747265+00:00"}