{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:DI7XWAMITIXKYZ4BR27OUQI62F","short_pith_number":"pith:DI7XWAMI","schema_version":"1.0","canonical_sha256":"1a3f7b01889a2eac67818ebeea411ed17de38b6f14d890db620483f95671ceae","source":{"kind":"arxiv","id":"2603.06591","version":2},"attestation_state":"computed","paper":{"title":"What Makes Position Zero Special? A Mechanistic Study of Position Zero Attention Sinks in LLMs","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.LG","authors_text":"Chen Zhao, Mingshu Chen, Qipeng Guo, Ruixiao Li, Runyu Peng, Xipeng Qiu, Yucheng Lu, Yunhua Zhou","submitted_at":"2026-02-04T11:10:34Z","abstract_excerpt":"Transformers frequently allocate disproportionate attention to specific tokens, a phenomenon known as attention sinks. Causal large language models reliably form one at position zero, though its role remains debated. We approach this question from a mechanistic perspective, tracing how the position-zero sink arises from the model's internal computation. We identify a two-block subnetwork responsible for this behavior, which we term the P0-Sink Circuit, and show it arises purely from the structural properties of causal attention, requiring no semantic content. We further validate through from-s"},"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":"2603.06591","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-02-04T11:10:34Z","cross_cats_sorted":["cs.CL"],"title_canon_sha256":"9001e01ba6117e5ce4d2b33e2a685f2cbfcdfa39e67ce5f506568e31c7ed6aac","abstract_canon_sha256":"a87cb490c1f37165ab633c2eb0025e2ed56b631a855aea99412c25ee6dda5d6e"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-08-04T01:59:35.058425Z","signature_b64":"lLMKHuKN8j9YMNV3ZzDWspe4guCtvCaEvo2IwNFLjCdtYpqg2DBbUqhNXpqHWM1SC1xO/Zs4VRAk/uDMq7G2Ag==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"1a3f7b01889a2eac67818ebeea411ed17de38b6f14d890db620483f95671ceae","last_reissued_at":"2026-08-04T01:59:35.056762Z","signature_status":"signed_v1","first_computed_at":"2026-08-04T01:59:35.056762Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"What Makes Position Zero Special? A Mechanistic Study of Position Zero Attention Sinks in LLMs","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.LG","authors_text":"Chen Zhao, Mingshu Chen, Qipeng Guo, Ruixiao Li, Runyu Peng, Xipeng Qiu, Yucheng Lu, Yunhua Zhou","submitted_at":"2026-02-04T11:10:34Z","abstract_excerpt":"Transformers frequently allocate disproportionate attention to specific tokens, a phenomenon known as attention sinks. Causal large language models reliably form one at position zero, though its role remains debated. We approach this question from a mechanistic perspective, tracing how the position-zero sink arises from the model's internal computation. We identify a two-block subnetwork responsible for this behavior, which we term the P0-Sink Circuit, and show it arises purely from the structural properties of causal attention, requiring no semantic content. We further validate through from-s"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2603.06591","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":""},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2603.06591/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":"2603.06591","created_at":"2026-08-04T01:59:35.058143+00:00"},{"alias_kind":"arxiv_version","alias_value":"2603.06591v2","created_at":"2026-08-04T01:59:35.058143+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2603.06591","created_at":"2026-08-04T01:59:35.058143+00:00"},{"alias_kind":"pith_short_12","alias_value":"DI7XWAMITIXK","created_at":"2026-08-04T01:59:35.058143+00:00"},{"alias_kind":"pith_short_16","alias_value":"DI7XWAMITIXKYZ4B","created_at":"2026-08-04T01:59:35.058143+00:00"},{"alias_kind":"pith_short_8","alias_value":"DI7XWAMI","created_at":"2026-08-04T01:59:35.058143+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":2,"internal_anchor_count":2,"sample":[{"citing_arxiv_id":"2606.02575","citing_title":"From Zero to Hero: Training-Free Custom Concept Spawning in World Models","ref_index":24,"is_internal_anchor":true},{"citing_arxiv_id":"2605.05696","citing_title":"Irminsul: MLA-Native Position-Independent Caching for Agentic LLM Serving","ref_index":26,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/DI7XWAMITIXKYZ4BR27OUQI62F","json":"https://pith.science/pith/DI7XWAMITIXKYZ4BR27OUQI62F.json","graph_json":"https://pith.science/api/pith-number/DI7XWAMITIXKYZ4BR27OUQI62F/graph.json","events_json":"https://pith.science/api/pith-number/DI7XWAMITIXKYZ4BR27OUQI62F/events.json","paper":"https://pith.science/paper/DI7XWAMI"},"agent_actions":{"view_html":"https://pith.science/pith/DI7XWAMITIXKYZ4BR27OUQI62F","download_json":"https://pith.science/pith/DI7XWAMITIXKYZ4BR27OUQI62F.json","view_paper":"https://pith.science/paper/DI7XWAMI","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2603.06591&json=true","fetch_graph":"https://pith.science/api/pith-number/DI7XWAMITIXKYZ4BR27OUQI62F/graph.json","fetch_events":"https://pith.science/api/pith-number/DI7XWAMITIXKYZ4BR27OUQI62F/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/DI7XWAMITIXKYZ4BR27OUQI62F/action/timestamp_anchor","attest_storage":"https://pith.science/pith/DI7XWAMITIXKYZ4BR27OUQI62F/action/storage_attestation","attest_author":"https://pith.science/pith/DI7XWAMITIXKYZ4BR27OUQI62F/action/author_attestation","sign_citation":"https://pith.science/pith/DI7XWAMITIXKYZ4BR27OUQI62F/action/citation_signature","submit_replication":"https://pith.science/pith/DI7XWAMITIXKYZ4BR27OUQI62F/action/replication_record"}},"created_at":"2026-08-04T01:59:35.058143+00:00","updated_at":"2026-08-04T01:59:35.058143+00:00"}