{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:YTXR37YDXGHVW2T7FKDAW4SFFG","short_pith_number":"pith:YTXR37YD","schema_version":"1.0","canonical_sha256":"c4ef1dff03b98f5b6a7f2a860b724529a8dac94a7e79ad61b45c42e37d465784","source":{"kind":"arxiv","id":"2505.20438","version":1},"attestation_state":"computed","paper":{"title":"HAMburger: Accelerating LLM Inference via Token Smashing","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Ce Zhang, Jingyu Liu","submitted_at":"2025-05-26T18:34:07Z","abstract_excerpt":"The growing demand for efficient Large Language Model (LLM) inference requires a holistic optimization on algorithms, systems, and hardware. However, very few works have fundamentally changed the generation pattern: each token needs one forward pass and one KV cache. This can be sub-optimal because we found that LLMs are extremely capable of self-identifying the exact dose of information that a single KV cache can store, and many tokens can be generated confidently without global context. Based on this insight, we introduce HAMburger, a Hierarchically Auto-regressive Model that redefines resou"},"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":"2505.20438","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2025-05-26T18:34:07Z","cross_cats_sorted":[],"title_canon_sha256":"b8a8a1858644de016d40f0aabe6e0729c90d4a0e37a3f675fa7ea8be4fa197bc","abstract_canon_sha256":"c0ee322a7a94c130769336ffd35b8315266a35e08d8e51d344223e51de6e2bea"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:10:12.501938Z","signature_b64":"Ug6OSxEWxpRLIHYAzK3r7U3dDN5MfGV7qJwydjPDv7MklveDs57tDWlioSL17uzd+8IlK6rP/H6uoELBeZK4CQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"c4ef1dff03b98f5b6a7f2a860b724529a8dac94a7e79ad61b45c42e37d465784","last_reissued_at":"2026-07-05T11:10:12.501397Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:10:12.501397Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"HAMburger: Accelerating LLM Inference via Token Smashing","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Ce Zhang, Jingyu Liu","submitted_at":"2025-05-26T18:34:07Z","abstract_excerpt":"The growing demand for efficient Large Language Model (LLM) inference requires a holistic optimization on algorithms, systems, and hardware. However, very few works have fundamentally changed the generation pattern: each token needs one forward pass and one KV cache. This can be sub-optimal because we found that LLMs are extremely capable of self-identifying the exact dose of information that a single KV cache can store, and many tokens can be generated confidently without global context. Based on this insight, we introduce HAMburger, a Hierarchically Auto-regressive Model that redefines resou"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2505.20438","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/2505.20438/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":"2505.20438","created_at":"2026-07-05T11:10:12.501455+00:00"},{"alias_kind":"arxiv_version","alias_value":"2505.20438v1","created_at":"2026-07-05T11:10:12.501455+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2505.20438","created_at":"2026-07-05T11:10:12.501455+00:00"},{"alias_kind":"pith_short_12","alias_value":"YTXR37YDXGHV","created_at":"2026-07-05T11:10:12.501455+00:00"},{"alias_kind":"pith_short_16","alias_value":"YTXR37YDXGHVW2T7","created_at":"2026-07-05T11:10:12.501455+00:00"},{"alias_kind":"pith_short_8","alias_value":"YTXR37YD","created_at":"2026-07-05T11:10:12.501455+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2507.16731","citing_title":"Collaborative Inference and Learning between Edge SLMs and Cloud LLMs: A Survey of Algorithms, Execution, and Open Challenges","ref_index":136,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/YTXR37YDXGHVW2T7FKDAW4SFFG","json":"https://pith.science/pith/YTXR37YDXGHVW2T7FKDAW4SFFG.json","graph_json":"https://pith.science/api/pith-number/YTXR37YDXGHVW2T7FKDAW4SFFG/graph.json","events_json":"https://pith.science/api/pith-number/YTXR37YDXGHVW2T7FKDAW4SFFG/events.json","paper":"https://pith.science/paper/YTXR37YD"},"agent_actions":{"view_html":"https://pith.science/pith/YTXR37YDXGHVW2T7FKDAW4SFFG","download_json":"https://pith.science/pith/YTXR37YDXGHVW2T7FKDAW4SFFG.json","view_paper":"https://pith.science/paper/YTXR37YD","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2505.20438&json=true","fetch_graph":"https://pith.science/api/pith-number/YTXR37YDXGHVW2T7FKDAW4SFFG/graph.json","fetch_events":"https://pith.science/api/pith-number/YTXR37YDXGHVW2T7FKDAW4SFFG/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/YTXR37YDXGHVW2T7FKDAW4SFFG/action/timestamp_anchor","attest_storage":"https://pith.science/pith/YTXR37YDXGHVW2T7FKDAW4SFFG/action/storage_attestation","attest_author":"https://pith.science/pith/YTXR37YDXGHVW2T7FKDAW4SFFG/action/author_attestation","sign_citation":"https://pith.science/pith/YTXR37YDXGHVW2T7FKDAW4SFFG/action/citation_signature","submit_replication":"https://pith.science/pith/YTXR37YDXGHVW2T7FKDAW4SFFG/action/replication_record"}},"created_at":"2026-07-05T11:10:12.501455+00:00","updated_at":"2026-07-05T11:10:12.501455+00:00"}