{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2021:CU56NGINI5XLIVQHK4VFY4LBXU","short_pith_number":"pith:CU56NGIN","schema_version":"1.0","canonical_sha256":"153be6990d476eb45607572a5c7161bd08abf64500b55698c7a1d0d98a50cf49","source":{"kind":"arxiv","id":"2106.06087","version":3},"attestation_state":"computed","paper":{"title":"Causal Analysis of Syntactic Agreement Mechanisms in Neural Language Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Aaron Mueller, Matthew Finlayson, Sebastian Gehrmann, Stuart Shieber, Tal Linzen, Yonatan Belinkov","submitted_at":"2021-06-10T23:50:51Z","abstract_excerpt":"Targeted syntactic evaluations have demonstrated the ability of language models to perform subject-verb agreement given difficult contexts. To elucidate the mechanisms by which the models accomplish this behavior, this study applies causal mediation analysis to pre-trained neural language models. We investigate the magnitude of models' preferences for grammatical inflections, as well as whether neurons process subject-verb agreement similarly across sentences with different syntactic structures. We uncover similarities and differences across architectures and model sizes -- notably, that large"},"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":"2106.06087","kind":"arxiv","version":3},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2021-06-10T23:50:51Z","cross_cats_sorted":[],"title_canon_sha256":"2862ca3accf1bc3988dd2b5f318a13dca419a1cbcf3edb6b9c7a84d59ecbc6fa","abstract_canon_sha256":"1028b2fc513b4b2fbe4e8327bcd5bdcea6858f57a736beb92fa9d55314ffa05e"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T02:51:27.256027Z","signature_b64":"LPN1+7vz2zn1JTPzNXLcHkVMiPcj9F4YMKavQRCDOwiv7KBYIR9CzjhaSodJfEy/MPdP2AJIPhk+ma5ZkQqvCQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"153be6990d476eb45607572a5c7161bd08abf64500b55698c7a1d0d98a50cf49","last_reissued_at":"2026-07-05T02:51:27.255483Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T02:51:27.255483Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Causal Analysis of Syntactic Agreement Mechanisms in Neural Language Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Aaron Mueller, Matthew Finlayson, Sebastian Gehrmann, Stuart Shieber, Tal Linzen, Yonatan Belinkov","submitted_at":"2021-06-10T23:50:51Z","abstract_excerpt":"Targeted syntactic evaluations have demonstrated the ability of language models to perform subject-verb agreement given difficult contexts. To elucidate the mechanisms by which the models accomplish this behavior, this study applies causal mediation analysis to pre-trained neural language models. We investigate the magnitude of models' preferences for grammatical inflections, as well as whether neurons process subject-verb agreement similarly across sentences with different syntactic structures. We uncover similarities and differences across architectures and model sizes -- notably, that large"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2106.06087","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/2106.06087/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":"2106.06087","created_at":"2026-07-05T02:51:27.255541+00:00"},{"alias_kind":"arxiv_version","alias_value":"2106.06087v3","created_at":"2026-07-05T02:51:27.255541+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2106.06087","created_at":"2026-07-05T02:51:27.255541+00:00"},{"alias_kind":"pith_short_12","alias_value":"CU56NGINI5XL","created_at":"2026-07-05T02:51:27.255541+00:00"},{"alias_kind":"pith_short_16","alias_value":"CU56NGINI5XLIVQH","created_at":"2026-07-05T02:51:27.255541+00:00"},{"alias_kind":"pith_short_8","alias_value":"CU56NGIN","created_at":"2026-07-05T02:51:27.255541+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":3,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2304.05969","citing_title":"Localizing Model Behavior with Path Patching","ref_index":38,"is_internal_anchor":false},{"citing_arxiv_id":"2211.00593","citing_title":"Interpretability in the Wild: a Circuit for Indirect Object Identification in GPT-2 small","ref_index":46,"is_internal_anchor":false},{"citing_arxiv_id":"2211.00593","citing_title":"Interpretability in the Wild: a Circuit for Indirect Object Identification in GPT-2 small","ref_index":24,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/CU56NGINI5XLIVQHK4VFY4LBXU","json":"https://pith.science/pith/CU56NGINI5XLIVQHK4VFY4LBXU.json","graph_json":"https://pith.science/api/pith-number/CU56NGINI5XLIVQHK4VFY4LBXU/graph.json","events_json":"https://pith.science/api/pith-number/CU56NGINI5XLIVQHK4VFY4LBXU/events.json","paper":"https://pith.science/paper/CU56NGIN"},"agent_actions":{"view_html":"https://pith.science/pith/CU56NGINI5XLIVQHK4VFY4LBXU","download_json":"https://pith.science/pith/CU56NGINI5XLIVQHK4VFY4LBXU.json","view_paper":"https://pith.science/paper/CU56NGIN","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2106.06087&json=true","fetch_graph":"https://pith.science/api/pith-number/CU56NGINI5XLIVQHK4VFY4LBXU/graph.json","fetch_events":"https://pith.science/api/pith-number/CU56NGINI5XLIVQHK4VFY4LBXU/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/CU56NGINI5XLIVQHK4VFY4LBXU/action/timestamp_anchor","attest_storage":"https://pith.science/pith/CU56NGINI5XLIVQHK4VFY4LBXU/action/storage_attestation","attest_author":"https://pith.science/pith/CU56NGINI5XLIVQHK4VFY4LBXU/action/author_attestation","sign_citation":"https://pith.science/pith/CU56NGINI5XLIVQHK4VFY4LBXU/action/citation_signature","submit_replication":"https://pith.science/pith/CU56NGINI5XLIVQHK4VFY4LBXU/action/replication_record"}},"created_at":"2026-07-05T02:51:27.255541+00:00","updated_at":"2026-07-05T02:51:27.255541+00:00"}