{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2023:CIILBVITOPVRIYTIW4NFEKUSW6","short_pith_number":"pith:CIILBVIT","schema_version":"1.0","canonical_sha256":"1210b0d51373eb146268b71a522a92b79841f1169b181a57bd037e4cb5c439e6","source":{"kind":"arxiv","id":"2312.09230","version":1},"attestation_state":"computed","paper":{"title":"Successor Heads: Recurring, Interpretable Attention Heads In The Wild","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.CL"],"primary_cat":"cs.LG","authors_text":"Arthur Conmy, Euan Ong, George Ogden, Rhys Gould","submitted_at":"2023-12-14T18:55:47Z","abstract_excerpt":"In this work we present successor heads: attention heads that increment tokens with a natural ordering, such as numbers, months, and days. For example, successor heads increment 'Monday' into 'Tuesday'. We explain the successor head behavior with an approach rooted in mechanistic interpretability, the field that aims to explain how models complete tasks in human-understandable terms. Existing research in this area has found interpretable language model components in small toy models. However, results in toy models have not yet led to insights that explain the internals of frontier models and l"},"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":"2312.09230","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2023-12-14T18:55:47Z","cross_cats_sorted":["cs.AI","cs.CL"],"title_canon_sha256":"1886ad0869210503b58a00443010aafb18b0452c3e53f16fb84b0f49179b1cf0","abstract_canon_sha256":"84aadf2784d3b32434dfa78244c6d9d218476d548bccb8c107b9bc42a26e0cfa"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T07:24:16.243012Z","signature_b64":"Cb2Gc5MV22mOsCDnMiYO/a0JXk9/q0VdnWwlmS/ydVqlIX8weJewzz9E+OmOgXMWayC9IabL00w2ZdrCxLC+Cw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"1210b0d51373eb146268b71a522a92b79841f1169b181a57bd037e4cb5c439e6","last_reissued_at":"2026-07-05T07:24:16.242545Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T07:24:16.242545Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Successor Heads: Recurring, Interpretable Attention Heads In The Wild","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.CL"],"primary_cat":"cs.LG","authors_text":"Arthur Conmy, Euan Ong, George Ogden, Rhys Gould","submitted_at":"2023-12-14T18:55:47Z","abstract_excerpt":"In this work we present successor heads: attention heads that increment tokens with a natural ordering, such as numbers, months, and days. For example, successor heads increment 'Monday' into 'Tuesday'. We explain the successor head behavior with an approach rooted in mechanistic interpretability, the field that aims to explain how models complete tasks in human-understandable terms. Existing research in this area has found interpretable language model components in small toy models. However, results in toy models have not yet led to insights that explain the internals of frontier models and l"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2312.09230","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/2312.09230/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":"2312.09230","created_at":"2026-07-05T07:24:16.242599+00:00"},{"alias_kind":"arxiv_version","alias_value":"2312.09230v1","created_at":"2026-07-05T07:24:16.242599+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2312.09230","created_at":"2026-07-05T07:24:16.242599+00:00"},{"alias_kind":"pith_short_12","alias_value":"CIILBVITOPVR","created_at":"2026-07-05T07:24:16.242599+00:00"},{"alias_kind":"pith_short_16","alias_value":"CIILBVITOPVRIYTI","created_at":"2026-07-05T07:24:16.242599+00:00"},{"alias_kind":"pith_short_8","alias_value":"CIILBVIT","created_at":"2026-07-05T07:24:16.242599+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":3,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.05378","citing_title":"Pattern Selectivity is Not Task-Causal Structure: A Cross-Architecture Mechanistic Study of Composed-Task Circuits in 1B-Class Language Models","ref_index":10,"is_internal_anchor":false},{"citing_arxiv_id":"2506.02546","citing_title":"To trust or not to trust: Attention-based Trust Management for LLM Multi-Agent Systems","ref_index":28,"is_internal_anchor":false},{"citing_arxiv_id":"2403.19647","citing_title":"Sparse Feature Circuits: Discovering and Editing Interpretable Causal Graphs in Language Models","ref_index":26,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/CIILBVITOPVRIYTIW4NFEKUSW6","json":"https://pith.science/pith/CIILBVITOPVRIYTIW4NFEKUSW6.json","graph_json":"https://pith.science/api/pith-number/CIILBVITOPVRIYTIW4NFEKUSW6/graph.json","events_json":"https://pith.science/api/pith-number/CIILBVITOPVRIYTIW4NFEKUSW6/events.json","paper":"https://pith.science/paper/CIILBVIT"},"agent_actions":{"view_html":"https://pith.science/pith/CIILBVITOPVRIYTIW4NFEKUSW6","download_json":"https://pith.science/pith/CIILBVITOPVRIYTIW4NFEKUSW6.json","view_paper":"https://pith.science/paper/CIILBVIT","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2312.09230&json=true","fetch_graph":"https://pith.science/api/pith-number/CIILBVITOPVRIYTIW4NFEKUSW6/graph.json","fetch_events":"https://pith.science/api/pith-number/CIILBVITOPVRIYTIW4NFEKUSW6/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/CIILBVITOPVRIYTIW4NFEKUSW6/action/timestamp_anchor","attest_storage":"https://pith.science/pith/CIILBVITOPVRIYTIW4NFEKUSW6/action/storage_attestation","attest_author":"https://pith.science/pith/CIILBVITOPVRIYTIW4NFEKUSW6/action/author_attestation","sign_citation":"https://pith.science/pith/CIILBVITOPVRIYTIW4NFEKUSW6/action/citation_signature","submit_replication":"https://pith.science/pith/CIILBVITOPVRIYTIW4NFEKUSW6/action/replication_record"}},"created_at":"2026-07-05T07:24:16.242599+00:00","updated_at":"2026-07-05T07:24:16.242599+00:00"}