{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:D47I62X27SZQESLHFICPULUYAP","short_pith_number":"pith:D47I62X2","schema_version":"1.0","canonical_sha256":"1f3e8f6afafcb30249672a04fa2e9803c50141f37eecf8ad49b9f513f43f3963","source":{"kind":"arxiv","id":"2503.21929","version":2},"attestation_state":"computed","paper":{"title":"Local Normalization Distortion and the Thermodynamic Formalism of Decoding Strategies for Large Language Models","license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","headline":"","cross_cats":["cs.LG","math.DS"],"primary_cat":"cs.CL","authors_text":"Stuart Burrell, Tom Kempton","submitted_at":"2025-03-27T19:15:43Z","abstract_excerpt":"Advances in hardware and language model architecture have spurred a revolution in natural language generation. However, autoregressive models compute probability distributions over next-token choices, and sampling from these distributions, known as decoding, has received significantly less attention than other design choices. Existing decoding strategies are largely based on heuristics, resulting in methods that are difficult to apply or improve in a principled manner. We develop the theory of decoding strategies for language models by expressing popular decoding algorithms as equilibrium stat"},"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":"2503.21929","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","primary_cat":"cs.CL","submitted_at":"2025-03-27T19:15:43Z","cross_cats_sorted":["cs.LG","math.DS"],"title_canon_sha256":"b61ee933bf388dc70efa2b8c1a40cbf55e4ab3978ff2a4cf07750fce8b70be7a","abstract_canon_sha256":"7502dcff295bade2c1abdba9fa8b427d782aa3f30c72adb8d943a41105769a3e"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T12:07:05.696901Z","signature_b64":"ueecd4lRb4mWmCwhjSvJs/ku76Z5PwFJqkPuHOJNK1mQzQvkmXSFP2GRPZKCMeKLArTUMRBjJ5oiQ3pQHgMwAw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"1f3e8f6afafcb30249672a04fa2e9803c50141f37eecf8ad49b9f513f43f3963","last_reissued_at":"2026-07-05T12:07:05.696465Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T12:07:05.696465Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Local Normalization Distortion and the Thermodynamic Formalism of Decoding Strategies for Large Language Models","license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","headline":"","cross_cats":["cs.LG","math.DS"],"primary_cat":"cs.CL","authors_text":"Stuart Burrell, Tom Kempton","submitted_at":"2025-03-27T19:15:43Z","abstract_excerpt":"Advances in hardware and language model architecture have spurred a revolution in natural language generation. However, autoregressive models compute probability distributions over next-token choices, and sampling from these distributions, known as decoding, has received significantly less attention than other design choices. Existing decoding strategies are largely based on heuristics, resulting in methods that are difficult to apply or improve in a principled manner. We develop the theory of decoding strategies for language models by expressing popular decoding algorithms as equilibrium stat"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2503.21929","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/2503.21929/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":"2503.21929","created_at":"2026-07-05T12:07:05.696517+00:00"},{"alias_kind":"arxiv_version","alias_value":"2503.21929v2","created_at":"2026-07-05T12:07:05.696517+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2503.21929","created_at":"2026-07-05T12:07:05.696517+00:00"},{"alias_kind":"pith_short_12","alias_value":"D47I62X27SZQ","created_at":"2026-07-05T12:07:05.696517+00:00"},{"alias_kind":"pith_short_16","alias_value":"D47I62X27SZQESLH","created_at":"2026-07-05T12:07:05.696517+00:00"},{"alias_kind":"pith_short_8","alias_value":"D47I62X2","created_at":"2026-07-05T12:07:05.696517+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.27359","citing_title":"When are likely answers right? On Sequence Probability and Correctness in LLMs","ref_index":27,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/D47I62X27SZQESLHFICPULUYAP","json":"https://pith.science/pith/D47I62X27SZQESLHFICPULUYAP.json","graph_json":"https://pith.science/api/pith-number/D47I62X27SZQESLHFICPULUYAP/graph.json","events_json":"https://pith.science/api/pith-number/D47I62X27SZQESLHFICPULUYAP/events.json","paper":"https://pith.science/paper/D47I62X2"},"agent_actions":{"view_html":"https://pith.science/pith/D47I62X27SZQESLHFICPULUYAP","download_json":"https://pith.science/pith/D47I62X27SZQESLHFICPULUYAP.json","view_paper":"https://pith.science/paper/D47I62X2","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2503.21929&json=true","fetch_graph":"https://pith.science/api/pith-number/D47I62X27SZQESLHFICPULUYAP/graph.json","fetch_events":"https://pith.science/api/pith-number/D47I62X27SZQESLHFICPULUYAP/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/D47I62X27SZQESLHFICPULUYAP/action/timestamp_anchor","attest_storage":"https://pith.science/pith/D47I62X27SZQESLHFICPULUYAP/action/storage_attestation","attest_author":"https://pith.science/pith/D47I62X27SZQESLHFICPULUYAP/action/author_attestation","sign_citation":"https://pith.science/pith/D47I62X27SZQESLHFICPULUYAP/action/citation_signature","submit_replication":"https://pith.science/pith/D47I62X27SZQESLHFICPULUYAP/action/replication_record"}},"created_at":"2026-07-05T12:07:05.696517+00:00","updated_at":"2026-07-05T12:07:05.696517+00:00"}