{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2022:GKEBQHZOUDEA77XMILT44NLAZP","short_pith_number":"pith:GKEBQHZO","canonical_record":{"source":{"id":"2202.05262","kind":"arxiv","version":5},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2022-02-10T18:59:54Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"5ce9b4bbd2053dd244ea67114c94a80075b0edf61ce03c55d302752266c983ec","abstract_canon_sha256":"4b53bc8ae790872206209c4c8d74e42d44cc03ad4c0074f1bcba9a2de9b43519"},"schema_version":"1.0"},"canonical_sha256":"3288181f2ea0c80ffeec42e7ce3560cbc408fdfa9fc77f12611401d53aadf9a5","source":{"kind":"arxiv","id":"2202.05262","version":5},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2202.05262","created_at":"2026-05-17T23:38:48Z"},{"alias_kind":"arxiv_version","alias_value":"2202.05262v5","created_at":"2026-05-17T23:38:48Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2202.05262","created_at":"2026-05-17T23:38:48Z"},{"alias_kind":"pith_short_12","alias_value":"GKEBQHZOUDEA","created_at":"2026-05-18T12:33:33Z"},{"alias_kind":"pith_short_16","alias_value":"GKEBQHZOUDEA77XM","created_at":"2026-05-18T12:33:33Z"},{"alias_kind":"pith_short_8","alias_value":"GKEBQHZO","created_at":"2026-05-18T12:33:33Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2022:GKEBQHZOUDEA77XMILT44NLAZP","target":"record","payload":{"canonical_record":{"source":{"id":"2202.05262","kind":"arxiv","version":5},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2022-02-10T18:59:54Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"5ce9b4bbd2053dd244ea67114c94a80075b0edf61ce03c55d302752266c983ec","abstract_canon_sha256":"4b53bc8ae790872206209c4c8d74e42d44cc03ad4c0074f1bcba9a2de9b43519"},"schema_version":"1.0"},"canonical_sha256":"3288181f2ea0c80ffeec42e7ce3560cbc408fdfa9fc77f12611401d53aadf9a5","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-17T23:38:48.267670Z","signature_b64":"xKojHUdhm6ZrOnU62GNrO2+2iuFFXyXH7UPQjb1jcG5e9PIRz6jc6c3/ZIfSfQhAzWnf86CXL2bvLAN6GCn8Dg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"3288181f2ea0c80ffeec42e7ce3560cbc408fdfa9fc77f12611401d53aadf9a5","last_reissued_at":"2026-05-17T23:38:48.267193Z","signature_status":"signed_v1","first_computed_at":"2026-05-17T23:38:48.267193Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2202.05262","source_version":5,"attestation_state":"computed"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-05-17T23:38:48Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"XRYThoMiMIbBQJ5sIqbePeT3aZPjMMu7PRyvA+TcNoxevPoNglh49G12NuEVsnVZIn0zGeS77S92pX6127N+Dg==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-06-10T14:33:32.534478Z"},"content_sha256":"c38921fc6b75874cecfc78a771229304f3dcbfa3f847662b159e8bf389de9cd2","schema_version":"1.0","event_id":"sha256:c38921fc6b75874cecfc78a771229304f3dcbfa3f847662b159e8bf389de9cd2"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2022:GKEBQHZOUDEA77XMILT44NLAZP","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"Locating and Editing Factual Associations in GPT","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"Factual associations in GPT models are stored in localized mid-layer feed-forward computations that can be directly edited via rank-one weight updates.","cross_cats":["cs.LG"],"primary_cat":"cs.CL","authors_text":"Alex Andonian, David Bau, Kevin Meng, Yonatan Belinkov","submitted_at":"2022-02-10T18:59:54Z","abstract_excerpt":"We analyze the storage and recall of factual associations in autoregressive transformer language models, finding evidence that these associations correspond to localized, directly-editable computations. We first develop a causal intervention for identifying neuron activations that are decisive in a model's factual predictions. This reveals a distinct set of steps in middle-layer feed-forward modules that mediate factual predictions while processing subject tokens. To test our hypothesis that these computations correspond to factual association recall, we modify feed-forward weights to update s"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"We find that ROME is effective on a standard zero-shot relation extraction (zsRE) model-editing task, comparable to existing methods. To perform a more sensitive evaluation, we also evaluate ROME on a new dataset of counterfactual assertions, on which it simultaneously maintains both specificity and generalization, whereas other methods sacrifice one or another.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"The assumption that the causal intervention correctly isolates the decisive feed-forward computations for factual recall, and that a rank-one update to those weights changes the association without creating unmeasured side effects on the broader distribution of model behavior.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"Factual associations in autoregressive transformers are localized to mid-layer feed-forward modules and can be edited via rank-one model editing while preserving both specificity and generalization on counterfactual tests.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"Factual associations in GPT models are stored in localized mid-layer feed-forward computations that can be directly edited via rank-one weight updates.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"b2abb54c7d4c21749bfc35a611cee2fafeabc9c1546669f3b09b475630eed4ea"},"source":{"id":"2202.05262","kind":"arxiv","version":5},"verdict":{"id":"596fd162-bcc3-417a-9bd0-541a17654282","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-16T09:55:25.774121Z","strongest_claim":"We find that ROME is effective on a standard zero-shot relation extraction (zsRE) model-editing task, comparable to existing methods. To perform a more sensitive evaluation, we also evaluate ROME on a new dataset of counterfactual assertions, on which it simultaneously maintains both specificity and generalization, whereas other methods sacrifice one or another.","one_line_summary":"Factual associations in autoregressive transformers are localized to mid-layer feed-forward modules and can be edited via rank-one model editing while preserving both specificity and generalization on counterfactual tests.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"The assumption that the causal intervention correctly isolates the decisive feed-forward computations for factual recall, and that a rank-one update to those weights changes the association without creating unmeasured side effects on the broader distribution of model behavior.","pith_extraction_headline":"Factual associations in GPT models are stored in localized mid-layer feed-forward computations that can be directly edited via rank-one weight updates."},"references":{"count":43,"sample":[{"doi":"","year":2017,"title":"Fine-grained analysis of sentence embeddings using auxiliary prediction tasks","work_id":"b7bd2b01-1ceb-409d-9ec3-e841491e27b1","ref_index":1,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":1972,"title":"Anderson, J. A. A simple neural network generating an interactive memory. Mathematical biosciences, 14 0 (3-4): 0 197--220, 1972","work_id":"e6955ef9-c3c3-47d3-8a55-a361da59cac2","ref_index":2,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2020,"title":"Rewriting a deep generative model","work_id":"30b22db5-f3a2-4cd6-83dc-bb134d36c8b7","ref_index":3,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"10.1162/coli_a_00422","year":2021,"title":"Probing Classifiers: Promises, Shortcomings, and Advances","work_id":"3eabab74-ac71-4292-86ce-b0469cd4e6cf","ref_index":4,"cited_arxiv_id":"2102.12452","is_internal_anchor":true},{"doi":"10.1162/tacl_a_00254","year":2019,"title":"Analysis methods in neural language processing: A survey","work_id":"62f1a2a2-d87f-4f14-90cf-b4b293a4780c","ref_index":5,"cited_arxiv_id":"","is_internal_anchor":false}],"resolved_work":43,"snapshot_sha256":"73260a8da562501fd2f1f3f7a4d7f9b70137a4a1f7ce0d835bd03497c761cf52","internal_anchors":3},"formal_canon":{"evidence_count":2,"snapshot_sha256":"32953a508cd941293af0a1776cfdcfae878e118b70e9d4b2121fff6aef4acbd5"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"},"verdict_id":"596fd162-bcc3-417a-9bd0-541a17654282"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-05-17T23:38:48Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"6bSitkeR1vXP656Mf+mlXmDN7OMJvsT+rMNnafkmLLYIZ746nSGJ4/gUsAhl1NriYShs3FMSYqafCKQ8CMgUCA==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-06-10T14:33:32.534994Z"},"content_sha256":"85fd2b36dca2fe9a857b407d3761e6a264e2469e4366927f491aa8258e8be397","schema_version":"1.0","event_id":"sha256:85fd2b36dca2fe9a857b407d3761e6a264e2469e4366927f491aa8258e8be397"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/GKEBQHZOUDEA77XMILT44NLAZP/bundle.json","state_url":"https://pith.science/pith/GKEBQHZOUDEA77XMILT44NLAZP/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/GKEBQHZOUDEA77XMILT44NLAZP/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":"2026-06-10T14:33:32Z","links":{"resolver":"https://pith.science/pith/GKEBQHZOUDEA77XMILT44NLAZP","bundle":"https://pith.science/pith/GKEBQHZOUDEA77XMILT44NLAZP/bundle.json","state":"https://pith.science/pith/GKEBQHZOUDEA77XMILT44NLAZP/state.json","well_known_bundle":"https://pith.science/.well-known/pith/GKEBQHZOUDEA77XMILT44NLAZP/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2022:GKEBQHZOUDEA77XMILT44NLAZP","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":"4b53bc8ae790872206209c4c8d74e42d44cc03ad4c0074f1bcba9a2de9b43519","cross_cats_sorted":["cs.LG"],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2022-02-10T18:59:54Z","title_canon_sha256":"5ce9b4bbd2053dd244ea67114c94a80075b0edf61ce03c55d302752266c983ec"},"schema_version":"1.0","source":{"id":"2202.05262","kind":"arxiv","version":5}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2202.05262","created_at":"2026-05-17T23:38:48Z"},{"alias_kind":"arxiv_version","alias_value":"2202.05262v5","created_at":"2026-05-17T23:38:48Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2202.05262","created_at":"2026-05-17T23:38:48Z"},{"alias_kind":"pith_short_12","alias_value":"GKEBQHZOUDEA","created_at":"2026-05-18T12:33:33Z"},{"alias_kind":"pith_short_16","alias_value":"GKEBQHZOUDEA77XM","created_at":"2026-05-18T12:33:33Z"},{"alias_kind":"pith_short_8","alias_value":"GKEBQHZO","created_at":"2026-05-18T12:33:33Z"}],"graph_snapshots":[{"event_id":"sha256:85fd2b36dca2fe9a857b407d3761e6a264e2469e4366927f491aa8258e8be397","target":"graph","created_at":"2026-05-17T23:38:48Z","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":4,"items":[{"attestation":"unclaimed","claim_id":"C1","kind":"strongest_claim","source":"verdict.strongest_claim","status":"machine_extracted","text":"We find that ROME is effective on a standard zero-shot relation extraction (zsRE) model-editing task, comparable to existing methods. To perform a more sensitive evaluation, we also evaluate ROME on a new dataset of counterfactual assertions, on which it simultaneously maintains both specificity and generalization, whereas other methods sacrifice one or another."},{"attestation":"unclaimed","claim_id":"C2","kind":"weakest_assumption","source":"verdict.weakest_assumption","status":"machine_extracted","text":"The assumption that the causal intervention correctly isolates the decisive feed-forward computations for factual recall, and that a rank-one update to those weights changes the association without creating unmeasured side effects on the broader distribution of model behavior."},{"attestation":"unclaimed","claim_id":"C3","kind":"one_line_summary","source":"verdict.one_line_summary","status":"machine_extracted","text":"Factual associations in autoregressive transformers are localized to mid-layer feed-forward modules and can be edited via rank-one model editing while preserving both specificity and generalization on counterfactual tests."},{"attestation":"unclaimed","claim_id":"C4","kind":"headline","source":"verdict.pith_extraction.headline","status":"machine_extracted","text":"Factual associations in GPT models are stored in localized mid-layer feed-forward computations that can be directly edited via rank-one weight updates."}],"snapshot_sha256":"b2abb54c7d4c21749bfc35a611cee2fafeabc9c1546669f3b09b475630eed4ea"},"formal_canon":{"evidence_count":2,"snapshot_sha256":"32953a508cd941293af0a1776cfdcfae878e118b70e9d4b2121fff6aef4acbd5"},"paper":{"abstract_excerpt":"We analyze the storage and recall of factual associations in autoregressive transformer language models, finding evidence that these associations correspond to localized, directly-editable computations. We first develop a causal intervention for identifying neuron activations that are decisive in a model's factual predictions. This reveals a distinct set of steps in middle-layer feed-forward modules that mediate factual predictions while processing subject tokens. To test our hypothesis that these computations correspond to factual association recall, we modify feed-forward weights to update s","authors_text":"Alex Andonian, David Bau, Kevin Meng, Yonatan Belinkov","cross_cats":["cs.LG"],"headline":"Factual associations in GPT models are stored in localized mid-layer feed-forward computations that can be directly edited via rank-one weight updates.","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2022-02-10T18:59:54Z","title":"Locating and Editing Factual Associations in GPT"},"references":{"count":43,"internal_anchors":3,"resolved_work":43,"sample":[{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":1,"title":"Fine-grained analysis of sentence embeddings using auxiliary prediction tasks","work_id":"b7bd2b01-1ceb-409d-9ec3-e841491e27b1","year":2017},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":2,"title":"Anderson, J. A. A simple neural network generating an interactive memory. Mathematical biosciences, 14 0 (3-4): 0 197--220, 1972","work_id":"e6955ef9-c3c3-47d3-8a55-a361da59cac2","year":1972},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":3,"title":"Rewriting a deep generative model","work_id":"30b22db5-f3a2-4cd6-83dc-bb134d36c8b7","year":2020},{"cited_arxiv_id":"2102.12452","doi":"10.1162/coli_a_00422","is_internal_anchor":true,"ref_index":4,"title":"Probing Classifiers: Promises, Shortcomings, and Advances","work_id":"3eabab74-ac71-4292-86ce-b0469cd4e6cf","year":2021},{"cited_arxiv_id":"","doi":"10.1162/tacl_a_00254","is_internal_anchor":false,"ref_index":5,"title":"Analysis methods in neural language processing: A survey","work_id":"62f1a2a2-d87f-4f14-90cf-b4b293a4780c","year":2019}],"snapshot_sha256":"73260a8da562501fd2f1f3f7a4d7f9b70137a4a1f7ce0d835bd03497c761cf52"},"source":{"id":"2202.05262","kind":"arxiv","version":5},"verdict":{"created_at":"2026-05-16T09:55:25.774121Z","id":"596fd162-bcc3-417a-9bd0-541a17654282","model_set":{"reader":"grok-4.3"},"one_line_summary":"Factual associations in autoregressive transformers are localized to mid-layer feed-forward modules and can be edited via rank-one model editing while preserving both specificity and generalization on counterfactual tests.","pipeline_version":"pith-pipeline@v0.9.0","pith_extraction_headline":"Factual associations in GPT models are stored in localized mid-layer feed-forward computations that can be directly edited via rank-one weight updates.","strongest_claim":"We find that ROME is effective on a standard zero-shot relation extraction (zsRE) model-editing task, comparable to existing methods. To perform a more sensitive evaluation, we also evaluate ROME on a new dataset of counterfactual assertions, on which it simultaneously maintains both specificity and generalization, whereas other methods sacrifice one or another.","weakest_assumption":"The assumption that the causal intervention correctly isolates the decisive feed-forward computations for factual recall, and that a rank-one update to those weights changes the association without creating unmeasured side effects on the broader distribution of model behavior."}},"verdict_id":"596fd162-bcc3-417a-9bd0-541a17654282"}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:c38921fc6b75874cecfc78a771229304f3dcbfa3f847662b159e8bf389de9cd2","target":"record","created_at":"2026-05-17T23:38:48Z","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":"4b53bc8ae790872206209c4c8d74e42d44cc03ad4c0074f1bcba9a2de9b43519","cross_cats_sorted":["cs.LG"],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2022-02-10T18:59:54Z","title_canon_sha256":"5ce9b4bbd2053dd244ea67114c94a80075b0edf61ce03c55d302752266c983ec"},"schema_version":"1.0","source":{"id":"2202.05262","kind":"arxiv","version":5}},"canonical_sha256":"3288181f2ea0c80ffeec42e7ce3560cbc408fdfa9fc77f12611401d53aadf9a5","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"3288181f2ea0c80ffeec42e7ce3560cbc408fdfa9fc77f12611401d53aadf9a5","first_computed_at":"2026-05-17T23:38:48.267193Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-05-17T23:38:48.267193Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"xKojHUdhm6ZrOnU62GNrO2+2iuFFXyXH7UPQjb1jcG5e9PIRz6jc6c3/ZIfSfQhAzWnf86CXL2bvLAN6GCn8Dg==","signature_status":"signed_v1","signed_at":"2026-05-17T23:38:48.267670Z","signed_message":"canonical_sha256_bytes"},"source_id":"2202.05262","source_kind":"arxiv","source_version":5}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:c38921fc6b75874cecfc78a771229304f3dcbfa3f847662b159e8bf389de9cd2","sha256:85fd2b36dca2fe9a857b407d3761e6a264e2469e4366927f491aa8258e8be397"],"state_sha256":"f03b5521aecc2656361caefe350c5f5c21955a3bdd42647d5258d05ad446c90b"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"eqoMRilwnEGJ1vsZd+Nmy9DwcN/INInvrsYKEsgm1mHwC0unOm/DZw0Gid2JTBdlO0laLzMkRd/N6bqqEVHdBg==","signed_message":"bundle_sha256_bytes","signed_at":"2026-06-10T14:33:32.537356Z","bundle_sha256":"8c859d6eee2aba9ae50545b29475d293f31021fe4a93c1bbd74587be8ad1b901"}}