{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:KSNI7JP5WTK7FLA574YGKTFGCD","short_pith_number":"pith:KSNI7JP5","schema_version":"1.0","canonical_sha256":"549a8fa5fdb4d5f2ac1dff30654ca610f712b3ac1cbf67bbc778992136630a97","source":{"kind":"arxiv","id":"2509.08778","version":1},"attestation_state":"computed","paper":{"title":"Do All Autoregressive Transformers Remember Facts the Same Way? A Cross-Architecture Analysis of Recall Mechanisms","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Changho Seo, Haehyun Cho, Hyunil Kim, Minyeong Choe","submitted_at":"2025-09-10T17:06:55Z","abstract_excerpt":"Understanding how Transformer-based language models store and retrieve factual associations is critical for improving interpretability and enabling targeted model editing. Prior work, primarily on GPT-style models, has identified MLP modules in early layers as key contributors to factual recall. However, it remains unclear whether these findings generalize across different autoregressive architectures. To address this, we conduct a comprehensive evaluation of factual recall across several models -- including GPT, LLaMA, Qwen, and DeepSeek -- analyzing where and how factual information is encod"},"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":"2509.08778","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2025-09-10T17:06:55Z","cross_cats_sorted":[],"title_canon_sha256":"41c4d48899cfdcf18bbed55fde512f39bf88578eb8401fa77431025db937ebde","abstract_canon_sha256":"4bb4618e7b90ee9dfe6e24371259913e88c82be8c9b3b3b794e070a9add01d93"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T12:08:36.106140Z","signature_b64":"kh3PMCXlbAMLiafABu/giIJSxP4oeSGvBxSlBFvTgKwSpsyBrbt/ExBk8mDyZso8Nq5O2wMQEBhAdiLVRcqyBw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"549a8fa5fdb4d5f2ac1dff30654ca610f712b3ac1cbf67bbc778992136630a97","last_reissued_at":"2026-07-05T12:08:36.105545Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T12:08:36.105545Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Do All Autoregressive Transformers Remember Facts the Same Way? A Cross-Architecture Analysis of Recall Mechanisms","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Changho Seo, Haehyun Cho, Hyunil Kim, Minyeong Choe","submitted_at":"2025-09-10T17:06:55Z","abstract_excerpt":"Understanding how Transformer-based language models store and retrieve factual associations is critical for improving interpretability and enabling targeted model editing. Prior work, primarily on GPT-style models, has identified MLP modules in early layers as key contributors to factual recall. However, it remains unclear whether these findings generalize across different autoregressive architectures. To address this, we conduct a comprehensive evaluation of factual recall across several models -- including GPT, LLaMA, Qwen, and DeepSeek -- analyzing where and how factual information is encod"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2509.08778","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/2509.08778/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":"2509.08778","created_at":"2026-07-05T12:08:36.105607+00:00"},{"alias_kind":"arxiv_version","alias_value":"2509.08778v1","created_at":"2026-07-05T12:08:36.105607+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2509.08778","created_at":"2026-07-05T12:08:36.105607+00:00"},{"alias_kind":"pith_short_12","alias_value":"KSNI7JP5WTK7","created_at":"2026-07-05T12:08:36.105607+00:00"},{"alias_kind":"pith_short_16","alias_value":"KSNI7JP5WTK7FLA5","created_at":"2026-07-05T12:08:36.105607+00:00"},{"alias_kind":"pith_short_8","alias_value":"KSNI7JP5","created_at":"2026-07-05T12:08:36.105607+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/KSNI7JP5WTK7FLA574YGKTFGCD","json":"https://pith.science/pith/KSNI7JP5WTK7FLA574YGKTFGCD.json","graph_json":"https://pith.science/api/pith-number/KSNI7JP5WTK7FLA574YGKTFGCD/graph.json","events_json":"https://pith.science/api/pith-number/KSNI7JP5WTK7FLA574YGKTFGCD/events.json","paper":"https://pith.science/paper/KSNI7JP5"},"agent_actions":{"view_html":"https://pith.science/pith/KSNI7JP5WTK7FLA574YGKTFGCD","download_json":"https://pith.science/pith/KSNI7JP5WTK7FLA574YGKTFGCD.json","view_paper":"https://pith.science/paper/KSNI7JP5","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2509.08778&json=true","fetch_graph":"https://pith.science/api/pith-number/KSNI7JP5WTK7FLA574YGKTFGCD/graph.json","fetch_events":"https://pith.science/api/pith-number/KSNI7JP5WTK7FLA574YGKTFGCD/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/KSNI7JP5WTK7FLA574YGKTFGCD/action/timestamp_anchor","attest_storage":"https://pith.science/pith/KSNI7JP5WTK7FLA574YGKTFGCD/action/storage_attestation","attest_author":"https://pith.science/pith/KSNI7JP5WTK7FLA574YGKTFGCD/action/author_attestation","sign_citation":"https://pith.science/pith/KSNI7JP5WTK7FLA574YGKTFGCD/action/citation_signature","submit_replication":"https://pith.science/pith/KSNI7JP5WTK7FLA574YGKTFGCD/action/replication_record"}},"created_at":"2026-07-05T12:08:36.105607+00:00","updated_at":"2026-07-05T12:08:36.105607+00:00"}