{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2021:O5RQ5RNLSBK3HZ4BSBEH3774LZ","short_pith_number":"pith:O5RQ5RNL","schema_version":"1.0","canonical_sha256":"77630ec5ab9055b3e78190487dfffc5e79dde68ced3eca852fc4e556b8cea672","source":{"kind":"arxiv","id":"2104.02112","version":2},"attestation_state":"computed","paper":{"title":"Efficient Attentions for Long Document Summarization","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Heng Ji, Lu Wang, Luyang Huang, Nikolaus Parulian, Shuyang Cao","submitted_at":"2021-04-05T18:45:13Z","abstract_excerpt":"The quadratic computational and memory complexities of large Transformers have limited their scalability for long document summarization. In this paper, we propose Hepos, a novel efficient encoder-decoder attention with head-wise positional strides to effectively pinpoint salient information from the source. We further conduct a systematic study of existing efficient self-attentions. Combined with Hepos, we are able to process ten times more tokens than existing models that use full attentions. For evaluation, we present a new dataset, GovReport, with significantly longer documents and summari"},"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":"2104.02112","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2021-04-05T18:45:13Z","cross_cats_sorted":[],"title_canon_sha256":"feb344fcb23902a4eb395e9e35ccbea01280c07e265e99e99a2b6a4b7bf35604","abstract_canon_sha256":"ecfed84661c1ca80736b092f3e3f8fad667dffb3155c2fc67888cf7dc0b2114b"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T02:31:05.399539Z","signature_b64":"vGZ85SBza5HIc15JHLLyZ1eMnViKtHOhoEvSfviy9xf7vcdIoaHOA+6+RGrKeMZ6YmHEhDdB2xpSDD8sNNKcBg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"77630ec5ab9055b3e78190487dfffc5e79dde68ced3eca852fc4e556b8cea672","last_reissued_at":"2026-07-05T02:31:05.399035Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T02:31:05.399035Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Efficient Attentions for Long Document Summarization","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Heng Ji, Lu Wang, Luyang Huang, Nikolaus Parulian, Shuyang Cao","submitted_at":"2021-04-05T18:45:13Z","abstract_excerpt":"The quadratic computational and memory complexities of large Transformers have limited their scalability for long document summarization. In this paper, we propose Hepos, a novel efficient encoder-decoder attention with head-wise positional strides to effectively pinpoint salient information from the source. We further conduct a systematic study of existing efficient self-attentions. Combined with Hepos, we are able to process ten times more tokens than existing models that use full attentions. For evaluation, we present a new dataset, GovReport, with significantly longer documents and summari"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2104.02112","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/2104.02112/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":"2104.02112","created_at":"2026-07-05T02:31:05.399097+00:00"},{"alias_kind":"arxiv_version","alias_value":"2104.02112v2","created_at":"2026-07-05T02:31:05.399097+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2104.02112","created_at":"2026-07-05T02:31:05.399097+00:00"},{"alias_kind":"pith_short_12","alias_value":"O5RQ5RNLSBK3","created_at":"2026-07-05T02:31:05.399097+00:00"},{"alias_kind":"pith_short_16","alias_value":"O5RQ5RNLSBK3HZ4B","created_at":"2026-07-05T02:31:05.399097+00:00"},{"alias_kind":"pith_short_8","alias_value":"O5RQ5RNL","created_at":"2026-07-05T02:31:05.399097+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":12,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.07923","citing_title":"Larch: Learned Query Optimization for Semantic Predicates","ref_index":23,"is_internal_anchor":false},{"citing_arxiv_id":"2606.29708","citing_title":"Demystifying the Design Space and Best Practices for Heterogeneous LLM Inference and Serving","ref_index":19,"is_internal_anchor":false},{"citing_arxiv_id":"2606.29708","citing_title":"Demystifying the Design Space and Best Practices for Heterogeneous LLM Inference and Serving","ref_index":19,"is_internal_anchor":false},{"citing_arxiv_id":"2606.29563","citing_title":"Coverage-Driven KV Cache Eviction for Efficient and Improved Inference of LLM","ref_index":12,"is_internal_anchor":false},{"citing_arxiv_id":"2602.08686","citing_title":"CompilerKV: Risk-Adaptive KV Compression via Offline Experience Compilation","ref_index":10,"is_internal_anchor":false},{"citing_arxiv_id":"2407.11550","citing_title":"Ada-KV: Optimizing KV Cache Eviction by Adaptive Budget Allocation for Efficient LLM Inference","ref_index":56,"is_internal_anchor":false},{"citing_arxiv_id":"2604.28157","citing_title":"FlashRT: Towards Computationally and Memory Efficient Red-Teaming for Prompt Injection and Knowledge Corruption","ref_index":56,"is_internal_anchor":false},{"citing_arxiv_id":"2604.24647","citing_title":"DepthKV: Layer-Dependent KV Cache Pruning for Long-Context LLM Inference","ref_index":1,"is_internal_anchor":false},{"citing_arxiv_id":"2307.08621","citing_title":"Retentive Network: A Successor to Transformer for Large Language Models","ref_index":8,"is_internal_anchor":false},{"citing_arxiv_id":"2604.21231","citing_title":"SparKV: Overhead-Aware KV Cache Loading for Efficient On-Device LLM Inference","ref_index":45,"is_internal_anchor":false},{"citing_arxiv_id":"2605.06850","citing_title":"How to Compress KV Cache in RL Post-Training? Shadow Mask Distillation for Memory-Efficient Alignment","ref_index":20,"is_internal_anchor":false},{"citing_arxiv_id":"2605.07234","citing_title":"Reformulating KV Cache Eviction Problem for Long-Context LLM Inference","ref_index":22,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/O5RQ5RNLSBK3HZ4BSBEH3774LZ","json":"https://pith.science/pith/O5RQ5RNLSBK3HZ4BSBEH3774LZ.json","graph_json":"https://pith.science/api/pith-number/O5RQ5RNLSBK3HZ4BSBEH3774LZ/graph.json","events_json":"https://pith.science/api/pith-number/O5RQ5RNLSBK3HZ4BSBEH3774LZ/events.json","paper":"https://pith.science/paper/O5RQ5RNL"},"agent_actions":{"view_html":"https://pith.science/pith/O5RQ5RNLSBK3HZ4BSBEH3774LZ","download_json":"https://pith.science/pith/O5RQ5RNLSBK3HZ4BSBEH3774LZ.json","view_paper":"https://pith.science/paper/O5RQ5RNL","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2104.02112&json=true","fetch_graph":"https://pith.science/api/pith-number/O5RQ5RNLSBK3HZ4BSBEH3774LZ/graph.json","fetch_events":"https://pith.science/api/pith-number/O5RQ5RNLSBK3HZ4BSBEH3774LZ/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/O5RQ5RNLSBK3HZ4BSBEH3774LZ/action/timestamp_anchor","attest_storage":"https://pith.science/pith/O5RQ5RNLSBK3HZ4BSBEH3774LZ/action/storage_attestation","attest_author":"https://pith.science/pith/O5RQ5RNLSBK3HZ4BSBEH3774LZ/action/author_attestation","sign_citation":"https://pith.science/pith/O5RQ5RNLSBK3HZ4BSBEH3774LZ/action/citation_signature","submit_replication":"https://pith.science/pith/O5RQ5RNLSBK3HZ4BSBEH3774LZ/action/replication_record"}},"created_at":"2026-07-05T02:31:05.399097+00:00","updated_at":"2026-07-05T02:31:05.399097+00:00"}