{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:NQPQJSB76UDZVURRCBGP4FM6EE","short_pith_number":"pith:NQPQJSB7","schema_version":"1.0","canonical_sha256":"6c1f04c83ff5079ad231104cfe159e212eca9c6476cdecd19257a51b07c5b025","source":{"kind":"arxiv","id":"2603.06875","version":3},"attestation_state":"computed","paper":{"title":"Stochastic Attention via Langevin Dynamics on the Modern Hopfield Energy","license":"http://creativecommons.org/licenses/by/4.0/","headline":"Attention retrieval equals one gradient step on the modern Hopfield energy, so Langevin dynamics yields a training-free stochastic sampler governed by temperature.","cross_cats":["q-fin.CP"],"primary_cat":"cs.LG","authors_text":"Abdulrahman Alswaidan, Jeffrey D. Varner","submitted_at":"2026-03-06T20:50:30Z","abstract_excerpt":"Attention heads retrieve: given a query, they return a weighted average of stored values. We showed that this computation is one step of gradient descent on the modern Hopfield energy, and that Langevin sampling from the corresponding Boltzmann distribution yielded stochastic attention, a training-free sampler controlled by a single temperature parameter. Lowering the temperature gave exact retrieval; raising it gave open-ended generation. Because the energy gradient equals the attention map, no score network, training loop, or learned model was required, making the approach particularly suite"},"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":true,"formal_links_present":false},"canonical_record":{"source":{"id":"2603.06875","kind":"arxiv","version":3},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-03-06T20:50:30Z","cross_cats_sorted":["q-fin.CP"],"title_canon_sha256":"5cc7d70025a2d57d48439e22dc7af2837180328af88442e501d52e0b6b0df148","abstract_canon_sha256":"61bbb0501458eb3bb092e59a5ca21b49717f83eb314e64c7afcd4358bc03ef88"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-17T23:38:59.736244Z","signature_b64":"4vlwG0xDmjxNutpeKbRA166BBfgq4Thy3nA6Kh5H1b2N13sYASsYZAoiIxA3HjFpnYlnacN5m8dv2ZPJcXVhAg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"6c1f04c83ff5079ad231104cfe159e212eca9c6476cdecd19257a51b07c5b025","last_reissued_at":"2026-05-17T23:38:59.735476Z","signature_status":"signed_v1","first_computed_at":"2026-05-17T23:38:59.735476Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Stochastic Attention via Langevin Dynamics on the Modern Hopfield Energy","license":"http://creativecommons.org/licenses/by/4.0/","headline":"Attention retrieval equals one gradient step on the modern Hopfield energy, so Langevin dynamics yields a training-free stochastic sampler governed by temperature.","cross_cats":["q-fin.CP"],"primary_cat":"cs.LG","authors_text":"Abdulrahman Alswaidan, Jeffrey D. Varner","submitted_at":"2026-03-06T20:50:30Z","abstract_excerpt":"Attention heads retrieve: given a query, they return a weighted average of stored values. We showed that this computation is one step of gradient descent on the modern Hopfield energy, and that Langevin sampling from the corresponding Boltzmann distribution yielded stochastic attention, a training-free sampler controlled by a single temperature parameter. Lowering the temperature gave exact retrieval; raising it gave open-ended generation. Because the energy gradient equals the attention map, no score network, training loop, or learned model was required, making the approach particularly suite"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"We showed that this computation is one step of gradient descent on the modern Hopfield energy, and that Langevin sampling from the corresponding Boltzmann distribution yielded stochastic attention, a training-free sampler controlled by a single temperature parameter.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"The assumption that the energy gradient exactly equals the attention map, allowing direct application of Langevin dynamics to produce valid samples without further modeling or approximations.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"Langevin sampling on the modern Hopfield energy produces training-free stochastic attention that transitions from exact retrieval to generation as temperature rises, with an entropy inflection condition marking the shift.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"Attention retrieval equals one gradient step on the modern Hopfield energy, so Langevin dynamics yields a training-free stochastic sampler governed by temperature.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"3213e47681779feec3712b11a338593cbb1369cc8065681bf5172c40340d0e88"},"source":{"id":"2603.06875","kind":"arxiv","version":3},"verdict":{"id":"01fef833-bc52-4afd-9460-763c56de0db6","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-15T14:39:46.526338Z","strongest_claim":"We showed that this computation is one step of gradient descent on the modern Hopfield energy, and that Langevin sampling from the corresponding Boltzmann distribution yielded stochastic attention, a training-free sampler controlled by a single temperature parameter.","one_line_summary":"Langevin sampling on the modern Hopfield energy produces training-free stochastic attention that transitions from exact retrieval to generation as temperature rises, with an entropy inflection condition marking the shift.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"The assumption that the energy gradient exactly equals the attention map, allowing direct application of Langevin dynamics to produce valid samples without further modeling or approximations.","pith_extraction_headline":"Attention retrieval equals one gradient step on the modern Hopfield energy, so Langevin dynamics yields a training-free stochastic sampler governed by temperature."},"references":{"count":40,"sample":[{"doi":"","year":2017,"title":"Gomez, Łukasz Kaiser, and Illia Polosukhin","work_id":"72294c37-bfde-4462-9168-556ddc61b278","ref_index":1,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"10.1073/pnas.79.8.2554","year":1982,"title":"Proceedings of the National Academy of Sci- ences79(8), 2554–2558 (Apr 1982)","work_id":"9d4d5a8f-b4dd-4c9d-809f-b85935997dbe","ref_index":2,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2016,"title":"Dmitry Krotov and John J. Hopfield. Dense associative memory for pattern recognition. In Advances in Neural Information Processing Systems, volume 29. Curran Associates, Inc., 2016","work_id":"ddec3c61-52d6-4d66-a665-193f1058601d","ref_index":3,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2021,"title":"Large associative memory problem in neurobiology and machine learning","work_id":"f50c1861-704a-4174-814c-11917f7e5375","ref_index":4,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2021,"title":"Hopfield networks is all you need","work_id":"04e53692-c9c5-4d7c-9347-9a0d34991317","ref_index":5,"cited_arxiv_id":"","is_internal_anchor":false}],"resolved_work":40,"snapshot_sha256":"cb05b0ee6ffbf3ead91c7e08f0eabd0ebf73b7681ed4bc2183040bf1394d23b8","internal_anchors":1},"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":"2603.06875","created_at":"2026-05-17T23:38:59.735603+00:00"},{"alias_kind":"arxiv_version","alias_value":"2603.06875v3","created_at":"2026-05-17T23:38:59.735603+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2603.06875","created_at":"2026-05-17T23:38:59.735603+00:00"},{"alias_kind":"pith_short_12","alias_value":"NQPQJSB76UDZ","created_at":"2026-05-18T12:33:37.589309+00:00"},{"alias_kind":"pith_short_16","alias_value":"NQPQJSB76UDZVURR","created_at":"2026-05-18T12:33:37.589309+00:00"},{"alias_kind":"pith_short_8","alias_value":"NQPQJSB7","created_at":"2026-05-18T12:33:37.589309+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/NQPQJSB76UDZVURRCBGP4FM6EE","json":"https://pith.science/pith/NQPQJSB76UDZVURRCBGP4FM6EE.json","graph_json":"https://pith.science/api/pith-number/NQPQJSB76UDZVURRCBGP4FM6EE/graph.json","events_json":"https://pith.science/api/pith-number/NQPQJSB76UDZVURRCBGP4FM6EE/events.json","paper":"https://pith.science/paper/NQPQJSB7"},"agent_actions":{"view_html":"https://pith.science/pith/NQPQJSB76UDZVURRCBGP4FM6EE","download_json":"https://pith.science/pith/NQPQJSB76UDZVURRCBGP4FM6EE.json","view_paper":"https://pith.science/paper/NQPQJSB7","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2603.06875&json=true","fetch_graph":"https://pith.science/api/pith-number/NQPQJSB76UDZVURRCBGP4FM6EE/graph.json","fetch_events":"https://pith.science/api/pith-number/NQPQJSB76UDZVURRCBGP4FM6EE/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/NQPQJSB76UDZVURRCBGP4FM6EE/action/timestamp_anchor","attest_storage":"https://pith.science/pith/NQPQJSB76UDZVURRCBGP4FM6EE/action/storage_attestation","attest_author":"https://pith.science/pith/NQPQJSB76UDZVURRCBGP4FM6EE/action/author_attestation","sign_citation":"https://pith.science/pith/NQPQJSB76UDZVURRCBGP4FM6EE/action/citation_signature","submit_replication":"https://pith.science/pith/NQPQJSB76UDZVURRCBGP4FM6EE/action/replication_record"}},"created_at":"2026-05-17T23:38:59.735603+00:00","updated_at":"2026-05-17T23:38:59.735603+00:00"}