{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2022:2DZYQWNB3UDGUNMC65OMVQMZJX","short_pith_number":"pith:2DZYQWNB","schema_version":"1.0","canonical_sha256":"d0f38859a1dd066a3582f75ccac1994df35a3177c515a19903a8898ec51aec91","source":{"kind":"arxiv","id":"2211.05102","version":1},"attestation_state":"computed","paper":{"title":"Efficiently Scaling Transformer Inference","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.LG","authors_text":"Aakanksha Chowdhery, Anselm Levskaya, Jacob Devlin, James Bradbury, Jeff Dean, Jonathan Heek, Kefan Xiao, Reiner Pope, Shivani Agrawal, Sholto Douglas","submitted_at":"2022-11-09T18:50:38Z","abstract_excerpt":"We study the problem of efficient generative inference for Transformer models, in one of its most challenging settings: large deep models, with tight latency targets and long sequence lengths. Better understanding of the engineering tradeoffs for inference for large Transformer-based models is important as use cases of these models are growing rapidly throughout application areas. We develop a simple analytical model for inference efficiency to select the best multi-dimensional partitioning techniques optimized for TPU v4 slices based on the application requirements. We combine these with a su"},"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":"2211.05102","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2022-11-09T18:50:38Z","cross_cats_sorted":["cs.CL"],"title_canon_sha256":"c5f690119287e5cb9ae5d5a79ad9e824a129162c7f3ee9d5d395416b65508654","abstract_canon_sha256":"80e3f4355f564ab92df021a937952999983ac36475c018970e1febdaa0c380ce"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T05:14:45.760670Z","signature_b64":"DuBY+IWE0oDj+ONgKTB4QVEh9IUdh0t84UA882b4bMKbPFzFRwi6EjWjX8jvhLUgCKqdp/4K9gyPqEd6S2CjDQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"d0f38859a1dd066a3582f75ccac1994df35a3177c515a19903a8898ec51aec91","last_reissued_at":"2026-07-05T05:14:45.760175Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T05:14:45.760175Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Efficiently Scaling Transformer Inference","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.LG","authors_text":"Aakanksha Chowdhery, Anselm Levskaya, Jacob Devlin, James Bradbury, Jeff Dean, Jonathan Heek, Kefan Xiao, Reiner Pope, Shivani Agrawal, Sholto Douglas","submitted_at":"2022-11-09T18:50:38Z","abstract_excerpt":"We study the problem of efficient generative inference for Transformer models, in one of its most challenging settings: large deep models, with tight latency targets and long sequence lengths. Better understanding of the engineering tradeoffs for inference for large Transformer-based models is important as use cases of these models are growing rapidly throughout application areas. We develop a simple analytical model for inference efficiency to select the best multi-dimensional partitioning techniques optimized for TPU v4 slices based on the application requirements. We combine these with a su"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2211.05102","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/2211.05102/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":"2211.05102","created_at":"2026-07-05T05:14:45.760233+00:00"},{"alias_kind":"arxiv_version","alias_value":"2211.05102v1","created_at":"2026-07-05T05:14:45.760233+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2211.05102","created_at":"2026-07-05T05:14:45.760233+00:00"},{"alias_kind":"pith_short_12","alias_value":"2DZYQWNB3UDG","created_at":"2026-07-05T05:14:45.760233+00:00"},{"alias_kind":"pith_short_16","alias_value":"2DZYQWNB3UDGUNMC","created_at":"2026-07-05T05:14:45.760233+00:00"},{"alias_kind":"pith_short_8","alias_value":"2DZYQWNB","created_at":"2026-07-05T05:14:45.760233+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":24,"internal_anchor_count":2,"sample":[{"citing_arxiv_id":"2607.05876","citing_title":"Think Before You Grid-Search: Floor-First Triage for LLM Serving","ref_index":27,"is_internal_anchor":true},{"citing_arxiv_id":"2607.05876","citing_title":"Think Before You Grid-Search: Floor-First Triage for LLM Serving","ref_index":27,"is_internal_anchor":true},{"citing_arxiv_id":"2606.24033","citing_title":"RoPE-Aware Bit Allocation for KV-Cache Quantization","ref_index":27,"is_internal_anchor":false},{"citing_arxiv_id":"2606.23419","citing_title":"GRINQH: Graded Input-based Quantization Hierarchy for Efficient LLM Generation","ref_index":33,"is_internal_anchor":false},{"citing_arxiv_id":"2606.23406","citing_title":"HyperQuant: A Rate-Distortion-Optimal Quantization Pipeline for Large Language and Diffusion Models","ref_index":30,"is_internal_anchor":false},{"citing_arxiv_id":"2606.30877","citing_title":"A Systematic Approach to Multi-Agent AI from Advanced Regulatory Control Theory: Safe and Auditable LLM Operator Agents for Process Control","ref_index":33,"is_internal_anchor":false},{"citing_arxiv_id":"2605.30580","citing_title":"Speculative Decoding Across Languages","ref_index":18,"is_internal_anchor":false},{"citing_arxiv_id":"2504.10013","citing_title":"Training LLMs on HPC Systems: Best Practices from the OpenGPT-X Project","ref_index":44,"is_internal_anchor":false},{"citing_arxiv_id":"2603.15031","citing_title":"Attention Residuals","ref_index":40,"is_internal_anchor":false},{"citing_arxiv_id":"2510.08055","citing_title":"From Tokens to Layers: Redefining Stall-Free Scheduling for MoE Serving with Layered Prefill","ref_index":12,"is_internal_anchor":false},{"citing_arxiv_id":"2511.00413","citing_title":"Tree Training: Accelerating Agentic LLMs Training via Shared Prefix Reuse","ref_index":11,"is_internal_anchor":false},{"citing_arxiv_id":"2306.14048","citing_title":"H$_2$O: Heavy-Hitter Oracle for Efficient Generative Inference of Large Language Models","ref_index":5,"is_internal_anchor":false},{"citing_arxiv_id":"2604.02337","citing_title":"Generating Counterfactual Patient Timelines from Real-World Data","ref_index":18,"is_internal_anchor":false},{"citing_arxiv_id":"2604.09593","citing_title":"Benchmarking Compound AI Applications for Hardware-Software Co-Design","ref_index":23,"is_internal_anchor":false},{"citing_arxiv_id":"2401.10774","citing_title":"Medusa: Simple LLM Inference Acceleration Framework with Multiple Decoding Heads","ref_index":38,"is_internal_anchor":false},{"citing_arxiv_id":"2309.06180","citing_title":"Efficient Memory Management for Large Language Model Serving with PagedAttention","ref_index":44,"is_internal_anchor":false},{"citing_arxiv_id":"2604.27039","citing_title":"Length Value Model: Scalable Value Pretraining for Token-Level Length Modeling","ref_index":17,"is_internal_anchor":false},{"citing_arxiv_id":"2604.25975","citing_title":"Rethinking KV Cache Eviction via a Unified Information-Theoretic Objective","ref_index":22,"is_internal_anchor":false},{"citing_arxiv_id":"2305.14314","citing_title":"QLoRA: Efficient Finetuning of Quantized LLMs","ref_index":47,"is_internal_anchor":false},{"citing_arxiv_id":"2604.20021","citing_title":"Continuous Semantic Caching for Low-Cost LLM Serving","ref_index":23,"is_internal_anchor":false},{"citing_arxiv_id":"2604.11659","citing_title":"GPU Acceleration of Sparse Fully Homomorphic Encrypted DNNs","ref_index":26,"is_internal_anchor":false},{"citing_arxiv_id":"2305.13245","citing_title":"GQA: Training Generalized Multi-Query Transformer Models from Multi-Head Checkpoints","ref_index":54,"is_internal_anchor":false},{"citing_arxiv_id":"2309.17453","citing_title":"Efficient Streaming Language Models with Attention Sinks","ref_index":39,"is_internal_anchor":false},{"citing_arxiv_id":"2604.04750","citing_title":"DeepStack: Scalable and Accurate Design Space Exploration for Distributed 3D-Stacked AI Accelerators","ref_index":80,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/2DZYQWNB3UDGUNMC65OMVQMZJX","json":"https://pith.science/pith/2DZYQWNB3UDGUNMC65OMVQMZJX.json","graph_json":"https://pith.science/api/pith-number/2DZYQWNB3UDGUNMC65OMVQMZJX/graph.json","events_json":"https://pith.science/api/pith-number/2DZYQWNB3UDGUNMC65OMVQMZJX/events.json","paper":"https://pith.science/paper/2DZYQWNB"},"agent_actions":{"view_html":"https://pith.science/pith/2DZYQWNB3UDGUNMC65OMVQMZJX","download_json":"https://pith.science/pith/2DZYQWNB3UDGUNMC65OMVQMZJX.json","view_paper":"https://pith.science/paper/2DZYQWNB","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2211.05102&json=true","fetch_graph":"https://pith.science/api/pith-number/2DZYQWNB3UDGUNMC65OMVQMZJX/graph.json","fetch_events":"https://pith.science/api/pith-number/2DZYQWNB3UDGUNMC65OMVQMZJX/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/2DZYQWNB3UDGUNMC65OMVQMZJX/action/timestamp_anchor","attest_storage":"https://pith.science/pith/2DZYQWNB3UDGUNMC65OMVQMZJX/action/storage_attestation","attest_author":"https://pith.science/pith/2DZYQWNB3UDGUNMC65OMVQMZJX/action/author_attestation","sign_citation":"https://pith.science/pith/2DZYQWNB3UDGUNMC65OMVQMZJX/action/citation_signature","submit_replication":"https://pith.science/pith/2DZYQWNB3UDGUNMC65OMVQMZJX/action/replication_record"}},"created_at":"2026-07-05T05:14:45.760233+00:00","updated_at":"2026-07-05T05:14:45.760233+00:00"}