{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2026:LZ3D3G46VLETAWS6EXEPKPGQD6","merge_version":"pith-open-graph-merge-v1","event_count":2,"valid_event_count":2,"invalid_event_count":0,"equivocation_count":0,"current":{"canonical_record":{"metadata":{"abstract_canon_sha256":"c8eee8d2dfea4f8b5dcebcadbf84a54bb0b23d3731d813dcc701e3dc1b3884d1","cross_cats_sorted":["cs.LG","math.PR"],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"stat.ML","submitted_at":"2026-05-12T19:48:36Z","title_canon_sha256":"e7e7361a6b0dd8924ec2cb1a074f514b5271392991d61c7399ec9260db43d215"},"schema_version":"1.0","source":{"id":"2605.12697","kind":"arxiv","version":1}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2605.12697","created_at":"2026-05-18T03:09:49Z"},{"alias_kind":"arxiv_version","alias_value":"2605.12697v1","created_at":"2026-05-18T03:09:49Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.12697","created_at":"2026-05-18T03:09:49Z"},{"alias_kind":"pith_short_12","alias_value":"LZ3D3G46VLET","created_at":"2026-05-18T12:33:37Z"},{"alias_kind":"pith_short_16","alias_value":"LZ3D3G46VLETAWS6","created_at":"2026-05-18T12:33:37Z"},{"alias_kind":"pith_short_8","alias_value":"LZ3D3G46","created_at":"2026-05-18T12:33:37Z"}],"graph_snapshots":[{"event_id":"sha256:7646c8ae91e3a81e18de36515d5a07281241b2d39337511410548a64196019ea","target":"graph","created_at":"2026-05-18T03:09:49Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"graph_snapshot":{"author_claims":{"count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","strong_count":0},"builder_version":"pith-number-builder-2026-05-17-v1","claims":{"count":4,"items":[{"attestation":"unclaimed","claim_id":"C1","kind":"strongest_claim","source":"verdict.strongest_claim","status":"machine_extracted","text":"we define an upper-tail accumulation scale and prove that it gives the critical inverse-temperature scale for softmax concentration: below this scale, the top competitors remain unseparated, whereas above it, the attention entropy collapses."},{"attestation":"unclaimed","claim_id":"C2","kind":"weakest_assumption","source":"verdict.weakest_assumption","status":"machine_extracted","text":"That the gap-counting function N_n of each attention row fully determines the critical scale and that the attention-score distributions admit well-defined successive gaps from the maximum."},{"attestation":"unclaimed","claim_id":"C3","kind":"one_line_summary","source":"verdict.one_line_summary","status":"machine_extracted","text":"The upper-tail accumulation scale derived from the gap-counting function N_n sets the critical inverse temperature for softmax attention concentration, unifying prior conflicting laws as special cases of different N_n."},{"attestation":"unclaimed","claim_id":"C4","kind":"headline","source":"verdict.pith_extraction.headline","status":"machine_extracted","text":"The critical inverse-temperature scale for self-attention concentration is fixed by an upper-tail accumulation scale derived from the gap-counting function of each attention row."}],"snapshot_sha256":"af513fa9b8c116bca3bf7ede604e09971e8c2013a855088d9e39b9f009f27f80"},"formal_canon":{"evidence_count":3,"snapshot_sha256":"dab92689155357b0c3c51b06698660927100bb511d22f38635a9f2d1129f175c"},"paper":{"abstract_excerpt":"Length-dependent logit rescaling is widely used to stabilize long-context self-attention, but existing analyses and methods suggest conflicting inverse-temperature laws for the context length $n$, ranging from $(\\log n)^{1/2}$ to $\\log n$ and $(\\log n)^2$. We provide a general theory showing that the desirable scale is determined by the gap-counting function $N_n$ of each attention row. Counting how many competitors lie within each gap from the maximum, we define an upper-tail accumulation scale and prove that it gives the critical inverse-temperature scale for softmax concentration: below thi","authors_text":"Ryo Karakida, Tomohiro Hayase","cross_cats":["cs.LG","math.PR"],"headline":"The critical inverse-temperature scale for self-attention concentration is fixed by an upper-tail accumulation scale derived from the gap-counting function of each attention row.","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"stat.ML","submitted_at":"2026-05-12T19:48:36Z","title":"A Unified Framework for Critical Scaling of Inverse Temperature in Self-Attention"},"references":{"count":32,"internal_anchors":3,"resolved_work":32,"sample":[{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":1,"title":"Attention is all you need","work_id":"4c6ff6f5-f994-47e1-b9da-213c378d21b8","year":2017},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":2,"title":"Infinite attention: NNGP and NTK for deep attention networks","work_id":"5edec609-57d6-4fff-a106-35d1df33eb4b","year":2020},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":3,"title":"Infinite limits of multi-head transformer dynamics","work_id":"455a8000-ba63-4352-a5f6-6c98bb18e2d6","year":2024},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":4,"title":"Infinite-width limit of a single attention layer: Analysis via tensor programs","work_id":"5998f624-fb37-489c-a484-ada3b2168e32","year":2025},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":5,"title":"A mathematical perspective on transformers","work_id":"d081e886-f9fa-4d25-9355-b6f6d2e2af62","year":2025}],"snapshot_sha256":"e0407cc42983bdab86c360498ce54eddd54b63baa391db3ca6e5cd661a1b5a3c"},"source":{"id":"2605.12697","kind":"arxiv","version":1},"verdict":{"created_at":"2026-05-14T19:50:11.335422Z","id":"8cc0e077-44f0-4d3f-aca3-acba8844ef1e","model_set":{"reader":"grok-4.3"},"one_line_summary":"The upper-tail accumulation scale derived from the gap-counting function N_n sets the critical inverse temperature for softmax attention concentration, unifying prior conflicting laws as special cases of different N_n.","pipeline_version":"pith-pipeline@v0.9.0","pith_extraction_headline":"The critical inverse-temperature scale for self-attention concentration is fixed by an upper-tail accumulation scale derived from the gap-counting function of each attention row.","strongest_claim":"we define an upper-tail accumulation scale and prove that it gives the critical inverse-temperature scale for softmax concentration: below this scale, the top competitors remain unseparated, whereas above it, the attention entropy collapses.","weakest_assumption":"That the gap-counting function N_n of each attention row fully determines the critical scale and that the attention-score distributions admit well-defined successive gaps from the maximum."}},"verdict_id":"8cc0e077-44f0-4d3f-aca3-acba8844ef1e"}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:641ff9b8255b936fe21aa53775d72283d6933ec1ef9d13fb1db334f4baba6dea","target":"record","created_at":"2026-05-18T03:09:49Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"attestation_state":"computed","canonical_record":{"metadata":{"abstract_canon_sha256":"c8eee8d2dfea4f8b5dcebcadbf84a54bb0b23d3731d813dcc701e3dc1b3884d1","cross_cats_sorted":["cs.LG","math.PR"],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"stat.ML","submitted_at":"2026-05-12T19:48:36Z","title_canon_sha256":"e7e7361a6b0dd8924ec2cb1a074f514b5271392991d61c7399ec9260db43d215"},"schema_version":"1.0","source":{"id":"2605.12697","kind":"arxiv","version":1}},"canonical_sha256":"5e763d9b9eaac9305a5e25c8f53cd01fa6b455df36ea251d7facbc29d4bf7732","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"5e763d9b9eaac9305a5e25c8f53cd01fa6b455df36ea251d7facbc29d4bf7732","first_computed_at":"2026-05-18T03:09:49.753423Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-05-18T03:09:49.753423Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"BZDeJL62Fz6lGE2kGWotgg8+4+NaBpXyulbLLoJgk+fUwQz7LdiGNARdf1NrszGJIj/86dZA5GGV7DkH1hXoDg==","signature_status":"signed_v1","signed_at":"2026-05-18T03:09:49.754197Z","signed_message":"canonical_sha256_bytes"},"source_id":"2605.12697","source_kind":"arxiv","source_version":1}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:641ff9b8255b936fe21aa53775d72283d6933ec1ef9d13fb1db334f4baba6dea","sha256:7646c8ae91e3a81e18de36515d5a07281241b2d39337511410548a64196019ea"],"state_sha256":"9fe9545a091998a65411e7c57e2e96a55e272fd2c080ab78fe5985a363e17f1a"}