{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:KVY76GSC3PJGAIXFR6ZTUFFDHY","short_pith_number":"pith:KVY76GSC","schema_version":"1.0","canonical_sha256":"5571ff1a42dbd26022e58fb33a14a33e132ce5e2cb38105138e4110309fa0816","source":{"kind":"arxiv","id":"2606.07604","version":1},"attestation_state":"computed","paper":{"title":"Contribution Weights: A Geometrical Analysis of Self-Attention Transformers","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Harry Jake Cunningham, Nicola Muca Cirone","submitted_at":"2026-05-29T09:40:38Z","abstract_excerpt":"Analyzing attention weights has become a standard approach for interpreting the information flow of Large Language Models (LLMs). However, this approach has significant limitations as it neglects the geometric properties of the value vectors being aggregated. To address this gap, we introduce \\emph{Contribution Weights}, a projection-based metric that quantifies a token's influence by accounting for it's attention weight, value magnitude, and directional alignment with the layer output. We demonstrate that contribution weights provide a more faithful measure of token importance, consistently o"},"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":"2606.07604","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-05-29T09:40:38Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"de06f053050dd1558e806d2a0b7427089c5af105aa49348d88699928579101b6","abstract_canon_sha256":"dec9e4ecfc2def8bca1567033fe6771487c160e60672dbae62440869525c266a"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-09T00:04:44.615625Z","signature_b64":"MCOiww23Uk+Ae5voZGFAUfylrZXpWigldPJE+WuP6xl4eIFYWWEfFvCxeEPcsUvAkkCfyBbn2N40MOHf8+YkBg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"5571ff1a42dbd26022e58fb33a14a33e132ce5e2cb38105138e4110309fa0816","last_reissued_at":"2026-06-09T00:04:44.615042Z","signature_status":"signed_v1","first_computed_at":"2026-06-09T00:04:44.615042Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Contribution Weights: A Geometrical Analysis of Self-Attention Transformers","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Harry Jake Cunningham, Nicola Muca Cirone","submitted_at":"2026-05-29T09:40:38Z","abstract_excerpt":"Analyzing attention weights has become a standard approach for interpreting the information flow of Large Language Models (LLMs). However, this approach has significant limitations as it neglects the geometric properties of the value vectors being aggregated. To address this gap, we introduce \\emph{Contribution Weights}, a projection-based metric that quantifies a token's influence by accounting for it's attention weight, value magnitude, and directional alignment with the layer output. We demonstrate that contribution weights provide a more faithful measure of token importance, consistently o"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2606.07604","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/2606.07604/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":"2606.07604","created_at":"2026-06-09T00:04:44.615132+00:00"},{"alias_kind":"arxiv_version","alias_value":"2606.07604v1","created_at":"2026-06-09T00:04:44.615132+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2606.07604","created_at":"2026-06-09T00:04:44.615132+00:00"},{"alias_kind":"pith_short_12","alias_value":"KVY76GSC3PJG","created_at":"2026-06-09T00:04:44.615132+00:00"},{"alias_kind":"pith_short_16","alias_value":"KVY76GSC3PJGAIXF","created_at":"2026-06-09T00:04:44.615132+00:00"},{"alias_kind":"pith_short_8","alias_value":"KVY76GSC","created_at":"2026-06-09T00:04:44.615132+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/KVY76GSC3PJGAIXFR6ZTUFFDHY","json":"https://pith.science/pith/KVY76GSC3PJGAIXFR6ZTUFFDHY.json","graph_json":"https://pith.science/api/pith-number/KVY76GSC3PJGAIXFR6ZTUFFDHY/graph.json","events_json":"https://pith.science/api/pith-number/KVY76GSC3PJGAIXFR6ZTUFFDHY/events.json","paper":"https://pith.science/paper/KVY76GSC"},"agent_actions":{"view_html":"https://pith.science/pith/KVY76GSC3PJGAIXFR6ZTUFFDHY","download_json":"https://pith.science/pith/KVY76GSC3PJGAIXFR6ZTUFFDHY.json","view_paper":"https://pith.science/paper/KVY76GSC","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2606.07604&json=true","fetch_graph":"https://pith.science/api/pith-number/KVY76GSC3PJGAIXFR6ZTUFFDHY/graph.json","fetch_events":"https://pith.science/api/pith-number/KVY76GSC3PJGAIXFR6ZTUFFDHY/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/KVY76GSC3PJGAIXFR6ZTUFFDHY/action/timestamp_anchor","attest_storage":"https://pith.science/pith/KVY76GSC3PJGAIXFR6ZTUFFDHY/action/storage_attestation","attest_author":"https://pith.science/pith/KVY76GSC3PJGAIXFR6ZTUFFDHY/action/author_attestation","sign_citation":"https://pith.science/pith/KVY76GSC3PJGAIXFR6ZTUFFDHY/action/citation_signature","submit_replication":"https://pith.science/pith/KVY76GSC3PJGAIXFR6ZTUFFDHY/action/replication_record"}},"created_at":"2026-06-09T00:04:44.615132+00:00","updated_at":"2026-06-09T00:04:44.615132+00:00"}