{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:QKNOCXGU37VTEACJ427TX64VQJ","short_pith_number":"pith:QKNOCXGU","schema_version":"1.0","canonical_sha256":"829ae15cd4dfeb320049e6bf3bfb958245711b00ae6d3e6caf5e9ddf2edeb79a","source":{"kind":"arxiv","id":"2605.25619","version":1},"attestation_state":"computed","paper":{"title":"Analogies between Transformer Layers and Power Method","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Chenglong Li, Claudio Altafini","submitted_at":"2026-05-25T09:20:41Z","abstract_excerpt":"In the paper we show that there is an analogy between the operations occurring in a layer of a transformer (projections and layer normalizations, disregarding the feedforward neural network) and a step in the power method. Coherently with this analogy, we show that passing through a layer the tokens tend to be tilted towards the principal eigenvector of a matrix which is the product of the output and value weight matrices of that layer. In the special case of a transformer with shared weights (i.e., in which all layers have identical weights) then the alignment with this principal eigenvector "},"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":"2605.25619","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-05-25T09:20:41Z","cross_cats_sorted":[],"title_canon_sha256":"cf8685e42ca104bae609c8e1fb57f0117b33bf569eb4582eaf9382cbcdfe81cd","abstract_canon_sha256":"5fc309c6cd0cc449bcf48328e0aaea63fad79bdef9fbf1b9c0741f6f87e46541"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-26T02:04:46.738387Z","signature_b64":"O/J4WvX6UeWI1H+zsC+IP9vRERxehzyPohQLcEUz4be+CtnQ2XNObdihGOeVgnrkimZv+2RWdl+/Sos6ItENCg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"829ae15cd4dfeb320049e6bf3bfb958245711b00ae6d3e6caf5e9ddf2edeb79a","last_reissued_at":"2026-05-26T02:04:46.737673Z","signature_status":"signed_v1","first_computed_at":"2026-05-26T02:04:46.737673Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Analogies between Transformer Layers and Power Method","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Chenglong Li, Claudio Altafini","submitted_at":"2026-05-25T09:20:41Z","abstract_excerpt":"In the paper we show that there is an analogy between the operations occurring in a layer of a transformer (projections and layer normalizations, disregarding the feedforward neural network) and a step in the power method. Coherently with this analogy, we show that passing through a layer the tokens tend to be tilted towards the principal eigenvector of a matrix which is the product of the output and value weight matrices of that layer. In the special case of a transformer with shared weights (i.e., in which all layers have identical weights) then the alignment with this principal eigenvector "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2605.25619","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/2605.25619/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":"2605.25619","created_at":"2026-05-26T02:04:46.737796+00:00"},{"alias_kind":"arxiv_version","alias_value":"2605.25619v1","created_at":"2026-05-26T02:04:46.737796+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.25619","created_at":"2026-05-26T02:04:46.737796+00:00"},{"alias_kind":"pith_short_12","alias_value":"QKNOCXGU37VT","created_at":"2026-05-26T02:04:46.737796+00:00"},{"alias_kind":"pith_short_16","alias_value":"QKNOCXGU37VTEACJ","created_at":"2026-05-26T02:04:46.737796+00:00"},{"alias_kind":"pith_short_8","alias_value":"QKNOCXGU","created_at":"2026-05-26T02:04:46.737796+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/QKNOCXGU37VTEACJ427TX64VQJ","json":"https://pith.science/pith/QKNOCXGU37VTEACJ427TX64VQJ.json","graph_json":"https://pith.science/api/pith-number/QKNOCXGU37VTEACJ427TX64VQJ/graph.json","events_json":"https://pith.science/api/pith-number/QKNOCXGU37VTEACJ427TX64VQJ/events.json","paper":"https://pith.science/paper/QKNOCXGU"},"agent_actions":{"view_html":"https://pith.science/pith/QKNOCXGU37VTEACJ427TX64VQJ","download_json":"https://pith.science/pith/QKNOCXGU37VTEACJ427TX64VQJ.json","view_paper":"https://pith.science/paper/QKNOCXGU","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2605.25619&json=true","fetch_graph":"https://pith.science/api/pith-number/QKNOCXGU37VTEACJ427TX64VQJ/graph.json","fetch_events":"https://pith.science/api/pith-number/QKNOCXGU37VTEACJ427TX64VQJ/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/QKNOCXGU37VTEACJ427TX64VQJ/action/timestamp_anchor","attest_storage":"https://pith.science/pith/QKNOCXGU37VTEACJ427TX64VQJ/action/storage_attestation","attest_author":"https://pith.science/pith/QKNOCXGU37VTEACJ427TX64VQJ/action/author_attestation","sign_citation":"https://pith.science/pith/QKNOCXGU37VTEACJ427TX64VQJ/action/citation_signature","submit_replication":"https://pith.science/pith/QKNOCXGU37VTEACJ427TX64VQJ/action/replication_record"}},"created_at":"2026-05-26T02:04:46.737796+00:00","updated_at":"2026-05-26T02:04:46.737796+00:00"}