{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2023:6CJWDV566PEGUO7KCQNNKGVSJ7","short_pith_number":"pith:6CJWDV56","canonical_record":{"source":{"id":"2310.04625","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2023-10-06T23:37:24Z","cross_cats_sorted":["cs.AI","cs.CL"],"title_canon_sha256":"ba01c30432dfc13481f3deb393571df30be3696928656834d31450ebbc267acb","abstract_canon_sha256":"15ea11c590581ea5f5b8eecad5c3fa3fa9a2e181d19572e8409bbc50cda1784e"},"schema_version":"1.0"},"canonical_sha256":"f09361d7bef3c86a3bea141ad51ab24ff519717d527cb5e54970f51fa348ba2b","source":{"kind":"arxiv","id":"2310.04625","version":1},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2310.04625","created_at":"2026-07-05T06:58:20Z"},{"alias_kind":"arxiv_version","alias_value":"2310.04625v1","created_at":"2026-07-05T06:58:20Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2310.04625","created_at":"2026-07-05T06:58:20Z"},{"alias_kind":"pith_short_12","alias_value":"6CJWDV566PEG","created_at":"2026-07-05T06:58:20Z"},{"alias_kind":"pith_short_16","alias_value":"6CJWDV566PEGUO7K","created_at":"2026-07-05T06:58:20Z"},{"alias_kind":"pith_short_8","alias_value":"6CJWDV56","created_at":"2026-07-05T06:58:20Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2023:6CJWDV566PEGUO7KCQNNKGVSJ7","target":"record","payload":{"canonical_record":{"source":{"id":"2310.04625","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2023-10-06T23:37:24Z","cross_cats_sorted":["cs.AI","cs.CL"],"title_canon_sha256":"ba01c30432dfc13481f3deb393571df30be3696928656834d31450ebbc267acb","abstract_canon_sha256":"15ea11c590581ea5f5b8eecad5c3fa3fa9a2e181d19572e8409bbc50cda1784e"},"schema_version":"1.0"},"canonical_sha256":"f09361d7bef3c86a3bea141ad51ab24ff519717d527cb5e54970f51fa348ba2b","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T06:58:20.843699Z","signature_b64":"2YwyLSYqFAOv91td92M6FtOzRpYAlrqSE5gAxXS/zwq9eYNSxN5xrBDBv4vhJa4COYYxOCBoJlaf5xksfn9CBg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"f09361d7bef3c86a3bea141ad51ab24ff519717d527cb5e54970f51fa348ba2b","last_reissued_at":"2026-07-05T06:58:20.843286Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T06:58:20.843286Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2310.04625","source_version":1,"attestation_state":"computed"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-07-05T06:58:20Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"ZuAn+hdz+EG/bsJlh7RUQvOh0V9TLnDnoNYzXxC4xHqAAevz9RYX4m5DWpnKdHgfLjERA4eCQQGJkwAZBN0bDw==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-09T13:29:31.534946Z"},"content_sha256":"ae6301dda4b3bb4b9806d46c7461a15f77eb83ff2a922ca45c929589d75884fd","schema_version":"1.0","event_id":"sha256:ae6301dda4b3bb4b9806d46c7461a15f77eb83ff2a922ca45c929589d75884fd"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2023:6CJWDV566PEGUO7KCQNNKGVSJ7","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"Copy Suppression: Comprehensively Understanding an Attention Head","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.CL"],"primary_cat":"cs.LG","authors_text":"Arthur Conmy, Callum McDougall, Cody Rushing, Neel Nanda, Thomas McGrath","submitted_at":"2023-10-06T23:37:24Z","abstract_excerpt":"We present a single attention head in GPT-2 Small that has one main role across the entire training distribution. If components in earlier layers predict a certain token, and this token appears earlier in the context, the head suppresses it: we call this copy suppression. Attention Head 10.7 (L10H7) suppresses naive copying behavior which improves overall model calibration. This explains why multiple prior works studying certain narrow tasks found negative heads that systematically favored the wrong answer. We uncover the mechanism that the Negative Heads use for copy suppression with weights-"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2310.04625","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/2310.04625/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"},"verdict_id":null},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-07-05T06:58:20Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"jAUW+v+V0/RiPvLX9hmSwdmSV9nF2f890mo+mrGUMZlNHlauyfnGwztn7g4qvdETHmWhOk/V8DiEKQQ9Lmh/Ag==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-09T13:29:31.535463Z"},"content_sha256":"689160072fc85bf175959e43f371d8826a001c77eec3e19418c1da0aa09b245c","schema_version":"1.0","event_id":"sha256:689160072fc85bf175959e43f371d8826a001c77eec3e19418c1da0aa09b245c"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/6CJWDV566PEGUO7KCQNNKGVSJ7/bundle.json","state_url":"https://pith.science/pith/6CJWDV566PEGUO7KCQNNKGVSJ7/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/6CJWDV566PEGUO7KCQNNKGVSJ7/bundle.json","status":"primary"}],"public_keys":[{"key_id":"pith-v1-2026-05","algorithm":"ed25519","format":"raw","public_key_b64":"stVStoiQhXFxp4s2pdzPNoqVNBMojDU/fJ2db5S3CbM=","public_key_hex":"b2d552b68890857171a78b36a5dccf368a953413288c353f7c9d9d6f94b709b3","fingerprint_sha256_b32_first128bits":"RVFV5Z2OI2J3ZUO7ERDEBCYNKS","fingerprint_sha256_hex":"8d4b5ee74e4693bcd1df2446408b0d54","rotates_at":null,"url":"https://pith.science/pith-signing-key.json","notes":"Pith uses this Ed25519 key to sign canonical record SHA-256 digests. Verify with: ed25519_verify(public_key, message=canonical_sha256_bytes, signature=base64decode(signature_b64))."}],"merge_version":"pith-open-graph-merge-v1","built_at":"2026-08-09T13:29:31Z","links":{"resolver":"https://pith.science/pith/6CJWDV566PEGUO7KCQNNKGVSJ7","bundle":"https://pith.science/pith/6CJWDV566PEGUO7KCQNNKGVSJ7/bundle.json","state":"https://pith.science/pith/6CJWDV566PEGUO7KCQNNKGVSJ7/state.json","well_known_bundle":"https://pith.science/.well-known/pith/6CJWDV566PEGUO7KCQNNKGVSJ7/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2023:6CJWDV566PEGUO7KCQNNKGVSJ7","merge_version":"pith-open-graph-merge-v1","event_count":2,"valid_event_count":2,"invalid_event_count":0,"equivocation_count":0,"current":{"canonical_record":{"metadata":{"abstract_canon_sha256":"15ea11c590581ea5f5b8eecad5c3fa3fa9a2e181d19572e8409bbc50cda1784e","cross_cats_sorted":["cs.AI","cs.CL"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2023-10-06T23:37:24Z","title_canon_sha256":"ba01c30432dfc13481f3deb393571df30be3696928656834d31450ebbc267acb"},"schema_version":"1.0","source":{"id":"2310.04625","kind":"arxiv","version":1}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2310.04625","created_at":"2026-07-05T06:58:20Z"},{"alias_kind":"arxiv_version","alias_value":"2310.04625v1","created_at":"2026-07-05T06:58:20Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2310.04625","created_at":"2026-07-05T06:58:20Z"},{"alias_kind":"pith_short_12","alias_value":"6CJWDV566PEG","created_at":"2026-07-05T06:58:20Z"},{"alias_kind":"pith_short_16","alias_value":"6CJWDV566PEGUO7K","created_at":"2026-07-05T06:58:20Z"},{"alias_kind":"pith_short_8","alias_value":"6CJWDV56","created_at":"2026-07-05T06:58:20Z"}],"graph_snapshots":[{"event_id":"sha256:689160072fc85bf175959e43f371d8826a001c77eec3e19418c1da0aa09b245c","target":"graph","created_at":"2026-07-05T06:58:20Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"graph_snapshot":{"author_claims":{"count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","strong_count":0},"builder_version":"pith-number-builder-2026-05-17-v1","claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"integrity":{"available":true,"clean":true,"detectors_run":[],"endpoint":"/pith/2310.04625/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"We present a single attention head in GPT-2 Small that has one main role across the entire training distribution. If components in earlier layers predict a certain token, and this token appears earlier in the context, the head suppresses it: we call this copy suppression. Attention Head 10.7 (L10H7) suppresses naive copying behavior which improves overall model calibration. This explains why multiple prior works studying certain narrow tasks found negative heads that systematically favored the wrong answer. We uncover the mechanism that the Negative Heads use for copy suppression with weights-","authors_text":"Arthur Conmy, Callum McDougall, Cody Rushing, Neel Nanda, Thomas McGrath","cross_cats":["cs.AI","cs.CL"],"headline":"","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2023-10-06T23:37:24Z","title":"Copy Suppression: Comprehensively Understanding an Attention Head"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2310.04625","kind":"arxiv","version":1},"verdict":{"created_at":null,"id":null,"model_set":{},"one_line_summary":"","pipeline_version":null,"pith_extraction_headline":"","strongest_claim":"","weakest_assumption":""}},"verdict_id":null}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:ae6301dda4b3bb4b9806d46c7461a15f77eb83ff2a922ca45c929589d75884fd","target":"record","created_at":"2026-07-05T06:58:20Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"attestation_state":"computed","canonical_record":{"metadata":{"abstract_canon_sha256":"15ea11c590581ea5f5b8eecad5c3fa3fa9a2e181d19572e8409bbc50cda1784e","cross_cats_sorted":["cs.AI","cs.CL"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2023-10-06T23:37:24Z","title_canon_sha256":"ba01c30432dfc13481f3deb393571df30be3696928656834d31450ebbc267acb"},"schema_version":"1.0","source":{"id":"2310.04625","kind":"arxiv","version":1}},"canonical_sha256":"f09361d7bef3c86a3bea141ad51ab24ff519717d527cb5e54970f51fa348ba2b","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"f09361d7bef3c86a3bea141ad51ab24ff519717d527cb5e54970f51fa348ba2b","first_computed_at":"2026-07-05T06:58:20.843286Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-07-05T06:58:20.843286Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"2YwyLSYqFAOv91td92M6FtOzRpYAlrqSE5gAxXS/zwq9eYNSxN5xrBDBv4vhJa4COYYxOCBoJlaf5xksfn9CBg==","signature_status":"signed_v1","signed_at":"2026-07-05T06:58:20.843699Z","signed_message":"canonical_sha256_bytes"},"source_id":"2310.04625","source_kind":"arxiv","source_version":1}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:ae6301dda4b3bb4b9806d46c7461a15f77eb83ff2a922ca45c929589d75884fd","sha256:689160072fc85bf175959e43f371d8826a001c77eec3e19418c1da0aa09b245c"],"state_sha256":"bbb2d621ae3073be8dd7c313dba17c5ccf672895cb60795f461ce8d2b9711ff2"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"vGKikqzsxsxhiMDih8eAX4d377XveWB0Wm97SCFkbBlL20wR3i65PPYyZDs/SuqaHcGe9rpAv5YYEG/vLC6SDw==","signed_message":"bundle_sha256_bytes","signed_at":"2026-08-09T13:29:31.540865Z","bundle_sha256":"8bba58ff3a2c7ccc78889ffbd1e5a7e77c2a40002b6bae91d4b72e60f4bf20c6"}}