{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2019:5ZTLN3WA5STZQHC6VVLFHCP7EZ","short_pith_number":"pith:5ZTLN3WA","canonical_record":{"source":{"id":"1906.04284","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2019-06-07T13:58:49Z","cross_cats_sorted":["cs.LG","stat.ML"],"title_canon_sha256":"8eeafd9f63e820923c5b67a556ced550aeef57939d10499e97e63c89bb7d538e","abstract_canon_sha256":"d69fae2c12483da60cf2b6f747d2ab7cd8880202cbd983dee46b3ba6e1456232"},"schema_version":"1.0"},"canonical_sha256":"ee66b6eec0eca7981c5ead565389ff2655810031a5b94b637601de03046bafef","source":{"kind":"arxiv","id":"1906.04284","version":2},"source_aliases":[{"alias_kind":"arxiv","alias_value":"1906.04284","created_at":"2026-05-17T23:42:58Z"},{"alias_kind":"arxiv_version","alias_value":"1906.04284v2","created_at":"2026-05-17T23:42:58Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.1906.04284","created_at":"2026-05-17T23:42:58Z"},{"alias_kind":"pith_short_12","alias_value":"5ZTLN3WA5STZ","created_at":"2026-05-18T12:33:10Z"},{"alias_kind":"pith_short_16","alias_value":"5ZTLN3WA5STZQHC6","created_at":"2026-05-18T12:33:10Z"},{"alias_kind":"pith_short_8","alias_value":"5ZTLN3WA","created_at":"2026-05-18T12:33:10Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2019:5ZTLN3WA5STZQHC6VVLFHCP7EZ","target":"record","payload":{"canonical_record":{"source":{"id":"1906.04284","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2019-06-07T13:58:49Z","cross_cats_sorted":["cs.LG","stat.ML"],"title_canon_sha256":"8eeafd9f63e820923c5b67a556ced550aeef57939d10499e97e63c89bb7d538e","abstract_canon_sha256":"d69fae2c12483da60cf2b6f747d2ab7cd8880202cbd983dee46b3ba6e1456232"},"schema_version":"1.0"},"canonical_sha256":"ee66b6eec0eca7981c5ead565389ff2655810031a5b94b637601de03046bafef","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-17T23:42:58.032307Z","signature_b64":"iAfszis3fU6qxXIg4eqytPsERpd+MBElMbM7rI1Lgzg8FkBPvD7xfhotfCZCAYWOjsfY3LyLo0oFTk0HpAa6Ag==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"ee66b6eec0eca7981c5ead565389ff2655810031a5b94b637601de03046bafef","last_reissued_at":"2026-05-17T23:42:58.031617Z","signature_status":"signed_v1","first_computed_at":"2026-05-17T23:42:58.031617Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"1906.04284","source_version":2,"attestation_state":"computed"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-05-17T23:42:58Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"ykhZnHXpe/iEeM562/4OeUMOVt+MuUpjkRQMjhfBYC7U3AzFxxBjBWEPxX9+KXnx68Yg9qeohp2WIgTM+xD4CQ==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-06-01T20:00:18.763168Z"},"content_sha256":"30b1768fc1c5b5681d9c730b75cc4d0d0793307a929ab1e90b3561d0b05d6c5c","schema_version":"1.0","event_id":"sha256:30b1768fc1c5b5681d9c730b75cc4d0d0793307a929ab1e90b3561d0b05d6c5c"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2019:5ZTLN3WA5STZQHC6VVLFHCP7EZ","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"Analyzing the Structure of Attention in a Transformer Language Model","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.LG","stat.ML"],"primary_cat":"cs.CL","authors_text":"Jesse Vig, Yonatan Belinkov","submitted_at":"2019-06-07T13:58:49Z","abstract_excerpt":"The Transformer is a fully attention-based alternative to recurrent networks that has achieved state-of-the-art results across a range of NLP tasks. In this paper, we analyze the structure of attention in a Transformer language model, the GPT-2 small pretrained model. We visualize attention for individual instances and analyze the interaction between attention and syntax over a large corpus. We find that attention targets different parts of speech at different layer depths within the model, and that attention aligns with dependency relations most strongly in the middle layers. We also find tha"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"1906.04284","kind":"arxiv","version":2},"verdict":{"id":null,"model_set":{},"created_at":null,"strongest_claim":"","one_line_summary":"","pipeline_version":null,"weakest_assumption":"","pith_extraction_headline":""},"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":"2026-05-17T23:42:58Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"Vd7mOThXR+N5K61QT15ZRS4LcVbmGSy7aTZpgRVS9BVcKqBNDw/RyhhNt0qFettzbAieMBpyDYngA/wDWUdZAw==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-06-01T20:00:18.763536Z"},"content_sha256":"9426296e19d70e1e5c434bdfac6d372f81c5e54168c37c957d2872951fb85473","schema_version":"1.0","event_id":"sha256:9426296e19d70e1e5c434bdfac6d372f81c5e54168c37c957d2872951fb85473"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/5ZTLN3WA5STZQHC6VVLFHCP7EZ/bundle.json","state_url":"https://pith.science/pith/5ZTLN3WA5STZQHC6VVLFHCP7EZ/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/5ZTLN3WA5STZQHC6VVLFHCP7EZ/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":"2026-06-01T20:00:18Z","links":{"resolver":"https://pith.science/pith/5ZTLN3WA5STZQHC6VVLFHCP7EZ","bundle":"https://pith.science/pith/5ZTLN3WA5STZQHC6VVLFHCP7EZ/bundle.json","state":"https://pith.science/pith/5ZTLN3WA5STZQHC6VVLFHCP7EZ/state.json","well_known_bundle":"https://pith.science/.well-known/pith/5ZTLN3WA5STZQHC6VVLFHCP7EZ/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2019:5ZTLN3WA5STZQHC6VVLFHCP7EZ","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":"d69fae2c12483da60cf2b6f747d2ab7cd8880202cbd983dee46b3ba6e1456232","cross_cats_sorted":["cs.LG","stat.ML"],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2019-06-07T13:58:49Z","title_canon_sha256":"8eeafd9f63e820923c5b67a556ced550aeef57939d10499e97e63c89bb7d538e"},"schema_version":"1.0","source":{"id":"1906.04284","kind":"arxiv","version":2}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"1906.04284","created_at":"2026-05-17T23:42:58Z"},{"alias_kind":"arxiv_version","alias_value":"1906.04284v2","created_at":"2026-05-17T23:42:58Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.1906.04284","created_at":"2026-05-17T23:42:58Z"},{"alias_kind":"pith_short_12","alias_value":"5ZTLN3WA5STZ","created_at":"2026-05-18T12:33:10Z"},{"alias_kind":"pith_short_16","alias_value":"5ZTLN3WA5STZQHC6","created_at":"2026-05-18T12:33:10Z"},{"alias_kind":"pith_short_8","alias_value":"5ZTLN3WA","created_at":"2026-05-18T12:33:10Z"}],"graph_snapshots":[{"event_id":"sha256:9426296e19d70e1e5c434bdfac6d372f81c5e54168c37c957d2872951fb85473","target":"graph","created_at":"2026-05-17T23:42:58Z","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"},"paper":{"abstract_excerpt":"The Transformer is a fully attention-based alternative to recurrent networks that has achieved state-of-the-art results across a range of NLP tasks. In this paper, we analyze the structure of attention in a Transformer language model, the GPT-2 small pretrained model. We visualize attention for individual instances and analyze the interaction between attention and syntax over a large corpus. We find that attention targets different parts of speech at different layer depths within the model, and that attention aligns with dependency relations most strongly in the middle layers. We also find tha","authors_text":"Jesse Vig, Yonatan Belinkov","cross_cats":["cs.LG","stat.ML"],"headline":"","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2019-06-07T13:58:49Z","title":"Analyzing the Structure of Attention in a Transformer Language Model"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"1906.04284","kind":"arxiv","version":2},"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":"sha256:30b1768fc1c5b5681d9c730b75cc4d0d0793307a929ab1e90b3561d0b05d6c5c","target":"record","created_at":"2026-05-17T23:42:58Z","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":"d69fae2c12483da60cf2b6f747d2ab7cd8880202cbd983dee46b3ba6e1456232","cross_cats_sorted":["cs.LG","stat.ML"],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2019-06-07T13:58:49Z","title_canon_sha256":"8eeafd9f63e820923c5b67a556ced550aeef57939d10499e97e63c89bb7d538e"},"schema_version":"1.0","source":{"id":"1906.04284","kind":"arxiv","version":2}},"canonical_sha256":"ee66b6eec0eca7981c5ead565389ff2655810031a5b94b637601de03046bafef","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"ee66b6eec0eca7981c5ead565389ff2655810031a5b94b637601de03046bafef","first_computed_at":"2026-05-17T23:42:58.031617Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-05-17T23:42:58.031617Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"iAfszis3fU6qxXIg4eqytPsERpd+MBElMbM7rI1Lgzg8FkBPvD7xfhotfCZCAYWOjsfY3LyLo0oFTk0HpAa6Ag==","signature_status":"signed_v1","signed_at":"2026-05-17T23:42:58.032307Z","signed_message":"canonical_sha256_bytes"},"source_id":"1906.04284","source_kind":"arxiv","source_version":2}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:30b1768fc1c5b5681d9c730b75cc4d0d0793307a929ab1e90b3561d0b05d6c5c","sha256:9426296e19d70e1e5c434bdfac6d372f81c5e54168c37c957d2872951fb85473"],"state_sha256":"22b9729daa77121199bfd77d527ec84857bd8aeff810d581970da91c3e69ca58"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"X3/S3xXXdAE5C6rmHvJB8I+j5ZT+lWc93wvgytpHu/Kx0OzGNZnJS1LVERxrJlBBNhUn5wODWjhzdezP6q3fDQ==","signed_message":"bundle_sha256_bytes","signed_at":"2026-06-01T20:00:18.765434Z","bundle_sha256":"43002c120ec89337d32a5f73eeb4f72f91d546ee0d60c1a6f4dfc1432050251b"}}