{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:B5WA6BGSARH6457R4HMIGAPXDT","short_pith_number":"pith:B5WA6BGS","schema_version":"1.0","canonical_sha256":"0f6c0f04d2044fee77f1e1d88301f71cc78898af96082ac86e1773d3b42991d4","source":{"kind":"arxiv","id":"2401.06761","version":1},"attestation_state":"computed","paper":{"title":"APAR: LLMs Can Do Auto-Parallel Auto-Regressive Decoding","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Aohan Zeng, Bowen Wang, Jie Tang, Mingdao Liu, Peng Zhang, Yuxiao Dong","submitted_at":"2024-01-12T18:50:36Z","abstract_excerpt":"The massive adoption of large language models (LLMs) demands efficient deployment strategies. However, the auto-regressive decoding process, which is fundamental to how most LLMs generate text, poses challenges to achieve efficient serving. In this work, we introduce a parallel auto-regressive generation method. By instruct-tuning on general domain data that contains hierarchical structures, we enable LLMs to independently plan their generation process and perform auto-parallel auto-regressive (APAR) generation, significantly reducing the number of generation steps. APAR alone can achieve up t"},"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":"2401.06761","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2024-01-12T18:50:36Z","cross_cats_sorted":[],"title_canon_sha256":"8aec5782dba780a17e96af1265ed7696b15254a740212c987119825ebd87ab6b","abstract_canon_sha256":"69d0a5cb5e012e36206bec1da2e1e610660f93da07711599d030071bb1c39451"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T07:33:04.939978Z","signature_b64":"IH55/qruSIbluOYzvr090Kkr5tRCAd1ZWUiQWcaJFOzA9g9Dtq89c3Nw841MZ6qgrxv3FkK9zkSwlQybfUpaDA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"0f6c0f04d2044fee77f1e1d88301f71cc78898af96082ac86e1773d3b42991d4","last_reissued_at":"2026-07-05T07:33:04.939565Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T07:33:04.939565Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"APAR: LLMs Can Do Auto-Parallel Auto-Regressive Decoding","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Aohan Zeng, Bowen Wang, Jie Tang, Mingdao Liu, Peng Zhang, Yuxiao Dong","submitted_at":"2024-01-12T18:50:36Z","abstract_excerpt":"The massive adoption of large language models (LLMs) demands efficient deployment strategies. However, the auto-regressive decoding process, which is fundamental to how most LLMs generate text, poses challenges to achieve efficient serving. In this work, we introduce a parallel auto-regressive generation method. By instruct-tuning on general domain data that contains hierarchical structures, we enable LLMs to independently plan their generation process and perform auto-parallel auto-regressive (APAR) generation, significantly reducing the number of generation steps. APAR alone can achieve up t"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2401.06761","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/2401.06761/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":"2401.06761","created_at":"2026-07-05T07:33:04.939619+00:00"},{"alias_kind":"arxiv_version","alias_value":"2401.06761v1","created_at":"2026-07-05T07:33:04.939619+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2401.06761","created_at":"2026-07-05T07:33:04.939619+00:00"},{"alias_kind":"pith_short_12","alias_value":"B5WA6BGSARH6","created_at":"2026-07-05T07:33:04.939619+00:00"},{"alias_kind":"pith_short_16","alias_value":"B5WA6BGSARH6457R","created_at":"2026-07-05T07:33:04.939619+00:00"},{"alias_kind":"pith_short_8","alias_value":"B5WA6BGS","created_at":"2026-07-05T07:33:04.939619+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":4,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.18697","citing_title":"PopPy: Opportunistically Exploiting Parallelism in Python Compound AI Applications","ref_index":46,"is_internal_anchor":false},{"citing_arxiv_id":"2404.14294","citing_title":"A Survey on Efficient Inference for Large Language Models","ref_index":49,"is_internal_anchor":false},{"citing_arxiv_id":"2604.26209","citing_title":"Breaking the Autoregressive Chain: Hyper-Parallel Decoding for Efficient LLM-Based Attribute Value Extraction","ref_index":5,"is_internal_anchor":false},{"citing_arxiv_id":"2406.12793","citing_title":"ChatGLM: A Family of Large Language Models from GLM-130B to GLM-4 All Tools","ref_index":21,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/B5WA6BGSARH6457R4HMIGAPXDT","json":"https://pith.science/pith/B5WA6BGSARH6457R4HMIGAPXDT.json","graph_json":"https://pith.science/api/pith-number/B5WA6BGSARH6457R4HMIGAPXDT/graph.json","events_json":"https://pith.science/api/pith-number/B5WA6BGSARH6457R4HMIGAPXDT/events.json","paper":"https://pith.science/paper/B5WA6BGS"},"agent_actions":{"view_html":"https://pith.science/pith/B5WA6BGSARH6457R4HMIGAPXDT","download_json":"https://pith.science/pith/B5WA6BGSARH6457R4HMIGAPXDT.json","view_paper":"https://pith.science/paper/B5WA6BGS","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2401.06761&json=true","fetch_graph":"https://pith.science/api/pith-number/B5WA6BGSARH6457R4HMIGAPXDT/graph.json","fetch_events":"https://pith.science/api/pith-number/B5WA6BGSARH6457R4HMIGAPXDT/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/B5WA6BGSARH6457R4HMIGAPXDT/action/timestamp_anchor","attest_storage":"https://pith.science/pith/B5WA6BGSARH6457R4HMIGAPXDT/action/storage_attestation","attest_author":"https://pith.science/pith/B5WA6BGSARH6457R4HMIGAPXDT/action/author_attestation","sign_citation":"https://pith.science/pith/B5WA6BGSARH6457R4HMIGAPXDT/action/citation_signature","submit_replication":"https://pith.science/pith/B5WA6BGSARH6457R4HMIGAPXDT/action/replication_record"}},"created_at":"2026-07-05T07:33:04.939619+00:00","updated_at":"2026-07-05T07:33:04.939619+00:00"}