{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2020:TJZ4X5Z3IOQ6GU5FOPQ3XEQ5RN","short_pith_number":"pith:TJZ4X5Z3","schema_version":"1.0","canonical_sha256":"9a73cbf73b43a1e353a573e1bb921d8b536f35c1e1325dc4d516c765a5766f55","source":{"kind":"arxiv","id":"2006.04768","version":3},"attestation_state":"computed","paper":{"title":"Linformer: Self-Attention with Linear Complexity","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"Self-attention in transformers can be approximated by a low-rank matrix to reduce complexity to linear in sequence length.","cross_cats":["stat.ML"],"primary_cat":"cs.LG","authors_text":"Belinda Z. Li, Han Fang, Hao Ma, Madian Khabsa, Sinong Wang","submitted_at":"2020-06-08T17:37:52Z","abstract_excerpt":"Large transformer models have shown extraordinary success in achieving state-of-the-art results in many natural language processing applications. However, training and deploying these models can be prohibitively costly for long sequences, as the standard self-attention mechanism of the Transformer uses $O(n^2)$ time and space with respect to sequence length. In this paper, we demonstrate that the self-attention mechanism can be approximated by a low-rank matrix. We further exploit this finding to propose a new self-attention mechanism, which reduces the overall self-attention complexity from $"},"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":true,"formal_links_present":true},"canonical_record":{"source":{"id":"2006.04768","kind":"arxiv","version":3},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2020-06-08T17:37:52Z","cross_cats_sorted":["stat.ML"],"title_canon_sha256":"b5efd79fc27856c2c0e08c6b6be2d77bca3e4535ab2b0f02c37293802bc2e9ef","abstract_canon_sha256":"0556f2b940c7c5bbbba7a010eb8e5581eed12c0783236e58aa8dc6b3c4621650"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T01:09:59.491288Z","signature_b64":"syKotT7VnbzUtMf21B7xaSXdUQagMp6XavZNxxnQfYkcjI/IwWUiY0kUhIfyCYPKkk9BNaj0qOz7D8TYMTbmDg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"9a73cbf73b43a1e353a573e1bb921d8b536f35c1e1325dc4d516c765a5766f55","last_reissued_at":"2026-07-05T01:09:59.490874Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T01:09:59.490874Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Linformer: Self-Attention with Linear Complexity","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"Self-attention in transformers can be approximated by a low-rank matrix to reduce complexity to linear in sequence length.","cross_cats":["stat.ML"],"primary_cat":"cs.LG","authors_text":"Belinda Z. Li, Han Fang, Hao Ma, Madian Khabsa, Sinong Wang","submitted_at":"2020-06-08T17:37:52Z","abstract_excerpt":"Large transformer models have shown extraordinary success in achieving state-of-the-art results in many natural language processing applications. However, training and deploying these models can be prohibitively costly for long sequences, as the standard self-attention mechanism of the Transformer uses $O(n^2)$ time and space with respect to sequence length. In this paper, we demonstrate that the self-attention mechanism can be approximated by a low-rank matrix. We further exploit this finding to propose a new self-attention mechanism, which reduces the overall self-attention complexity from $"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"We demonstrate that the self-attention mechanism can be approximated by a low-rank matrix... The resulting linear transformer, the Linformer, performs on par with standard Transformer models, while being much more memory- and time-efficient.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That the low-rank projection matrices learned or fixed during training preserve enough information for downstream tasks across the full range of sequence lengths and domains the model will encounter.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"Linformer approximates self-attention with a low-rank projection to achieve O(n) time and space complexity while matching Transformer accuracy on standard NLP tasks.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"Self-attention in transformers can be approximated by a low-rank matrix to reduce complexity to linear in sequence length.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"a487fe0efcf41a99445598c202731d9fbab80e318271895095e67dae94cf6208"},"source":{"id":"2006.04768","kind":"arxiv","version":3},"verdict":{"id":"0a08b9c0-8e77-493b-8931-242caa882ea7","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-12T00:29:37.213350Z","strongest_claim":"We demonstrate that the self-attention mechanism can be approximated by a low-rank matrix... The resulting linear transformer, the Linformer, performs on par with standard Transformer models, while being much more memory- and time-efficient.","one_line_summary":"Linformer approximates self-attention with a low-rank projection to achieve O(n) time and space complexity while matching Transformer accuracy on standard NLP tasks.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That the low-rank projection matrices learned or fixed during training preserve enough information for downstream tasks across the full range of sequence lengths and domains the model will encounter.","pith_extraction_headline":"Self-attention in transformers can be approximated by a low-rank matrix to reduce complexity to linear in sequence length."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2006.04768/integrity.json","findings":[],"available":true,"detectors_run":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938"},"references":{"count":19,"sample":[{"doi":"","year":2004,"title":"Longformer: The Long-Document Transformer","work_id":"abea7a44-6668-4de7-aab6-f53a6e5aa088","ref_index":1,"cited_arxiv_id":"2004.05150","is_internal_anchor":true},{"doi":"","year":2005,"title":"Language Models are Few-Shot Learners","work_id":"214732c0-2edd-44a0-af9e-28184a2b8279","ref_index":2,"cited_arxiv_id":"2005.14165","is_internal_anchor":true},{"doi":"","year":null,"title":"Training Deep Nets with Sublinear Memory Cost","work_id":"f2c5c287-a500-40e4-a136-e7e3172db1d7","ref_index":3,"cited_arxiv_id":"1604.06174","is_internal_anchor":true},{"doi":"","year":1904,"title":"Generating Long Sequences with Sparse Transformers","work_id":"c5b81688-45ee-4a9a-b095-e6290f45cb6c","ref_index":4,"cited_arxiv_id":"1904.10509","is_internal_anchor":true},{"doi":"","year":2019,"title":"Bert: Pre-training of deep bidirectional transformers for language understanding","work_id":"693b70ad-3022-4615-938e-7752341ec181","ref_index":5,"cited_arxiv_id":"","is_internal_anchor":false}],"resolved_work":19,"snapshot_sha256":"d7ab771ae1b9b3bcd851ffe57f3a2ee32c9ae27ca7672a6d6b9c0d536b970ec6","internal_anchors":11},"formal_canon":{"evidence_count":3,"snapshot_sha256":"2b22cdf79b79855b10388867a1a43a9d35d290d2b3eb673140f21aa4e921a432"},"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":"2006.04768","created_at":"2026-07-05T01:09:59.490931+00:00"},{"alias_kind":"arxiv_version","alias_value":"2006.04768v3","created_at":"2026-07-05T01:09:59.490931+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2006.04768","created_at":"2026-07-05T01:09:59.490931+00:00"},{"alias_kind":"pith_short_12","alias_value":"TJZ4X5Z3IOQ6","created_at":"2026-07-05T01:09:59.490931+00:00"},{"alias_kind":"pith_short_16","alias_value":"TJZ4X5Z3IOQ6GU5F","created_at":"2026-07-05T01:09:59.490931+00:00"},{"alias_kind":"pith_short_8","alias_value":"TJZ4X5Z3","created_at":"2026-07-05T01:09:59.490931+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":115,"internal_anchor_count":115,"sample":[{"citing_arxiv_id":"2606.27372","citing_title":"DnA: Denoising Attention for Visual Tasks","ref_index":76,"is_internal_anchor":true},{"citing_arxiv_id":"2606.26749","citing_title":"Structure Before Collapse: Transient semantic geometry in next-token prediction","ref_index":192,"is_internal_anchor":true},{"citing_arxiv_id":"2606.21848","citing_title":"Keyless Attention: Value-Space Routing and Value-Only Caching for Efficient Transformers","ref_index":49,"is_internal_anchor":true},{"citing_arxiv_id":"2606.19993","citing_title":"Activation- and Influence-Aware Ranks (AIR): Function-Preserving SVD Compression for LLMs","ref_index":98,"is_internal_anchor":true},{"citing_arxiv_id":"2606.13338","citing_title":"Navigating the Safety-Fidelity Trade-off: Massive-Variate Time Series Forecasting for Power Systems via Probabilistic Scenarios","ref_index":44,"is_internal_anchor":true},{"citing_arxiv_id":"2606.12895","citing_title":"LongSpike: Fractional Order Spiking State Space Models for Efficient Long Sequence Learning","ref_index":11,"is_internal_anchor":true},{"citing_arxiv_id":"2606.12059","citing_title":"Attention by Synchronization in Coupled Oscillator Networks","ref_index":9,"is_internal_anchor":true},{"citing_arxiv_id":"2606.08804","citing_title":"Q-Delta: Beyond Key-Value Associative State Evolution","ref_index":87,"is_internal_anchor":true},{"citing_arxiv_id":"2606.08327","citing_title":"Chiaroscuro Attention: Spending Compute in the Dark","ref_index":9,"is_internal_anchor":true},{"citing_arxiv_id":"2606.06573","citing_title":"Multiscale POD of Transformer Attention Fields: Scale-Selective Analysis via Morlet Scalogram","ref_index":30,"is_internal_anchor":true},{"citing_arxiv_id":"2607.00902","citing_title":"MG-RWKV: Multi-Grained Context-Aware RWKV for Temporal Forgery Localization","ref_index":42,"is_internal_anchor":true},{"citing_arxiv_id":"2606.04833","citing_title":"Signed Dual Attention: Capturing Signed Dependencies in Time Series Forecasting","ref_index":16,"is_internal_anchor":true},{"citing_arxiv_id":"2606.28405","citing_title":"Enhancing Layer Interaction Using Key-Correlated Layer Attention","ref_index":9,"is_internal_anchor":true},{"citing_arxiv_id":"2606.30460","citing_title":"HSAP: A Hierarchical Sequence-aware Parallelism for Hybrid-Context Generative Models","ref_index":27,"is_internal_anchor":true},{"citing_arxiv_id":"2605.14913","citing_title":"Representative Attention For Vision Transformers","ref_index":13,"is_internal_anchor":true},{"citing_arxiv_id":"2606.00077","citing_title":"Improved Belief-Attention in Vision Task","ref_index":14,"is_internal_anchor":true},{"citing_arxiv_id":"2605.24518","citing_title":"Grammatically-Guided Sparse Attention for Efficient and Interpretable Transformers","ref_index":16,"is_internal_anchor":true},{"citing_arxiv_id":"2605.24930","citing_title":"H$^{2}$MT: Semantic Hierarchy-Aware Hierarchical Memory Transformer","ref_index":21,"is_internal_anchor":true},{"citing_arxiv_id":"2606.28929","citing_title":"Cybersecurity is the True Frontier for Generative AI Success or Failure","ref_index":59,"is_internal_anchor":true},{"citing_arxiv_id":"2606.30460","citing_title":"HSAP: A Hierarchical Sequence-aware Parallelism for Hybrid-Context Generative Models","ref_index":27,"is_internal_anchor":true},{"citing_arxiv_id":"2605.25655","citing_title":"Bandwidth-Aware LLM Inference on Heterogeneous Many-Core Supercomputers","ref_index":26,"is_internal_anchor":true},{"citing_arxiv_id":"2605.27686","citing_title":"Tensor Memory: Fixed-Size Recurrent State for Long-Horizon Transformers","ref_index":36,"is_internal_anchor":true},{"citing_arxiv_id":"2606.24253","citing_title":"TuringViT: Making SOTA Vision Transformers Accessible to All","ref_index":27,"is_internal_anchor":true},{"citing_arxiv_id":"2605.29954","citing_title":"SwInception -- Local Attention Meets Convolutions","ref_index":27,"is_internal_anchor":true},{"citing_arxiv_id":"2605.31559","citing_title":"Functional Attention: From Pairwise Affinities to Functional Correspondences","ref_index":13,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":3,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/TJZ4X5Z3IOQ6GU5FOPQ3XEQ5RN","json":"https://pith.science/pith/TJZ4X5Z3IOQ6GU5FOPQ3XEQ5RN.json","graph_json":"https://pith.science/api/pith-number/TJZ4X5Z3IOQ6GU5FOPQ3XEQ5RN/graph.json","events_json":"https://pith.science/api/pith-number/TJZ4X5Z3IOQ6GU5FOPQ3XEQ5RN/events.json","paper":"https://pith.science/paper/TJZ4X5Z3"},"agent_actions":{"view_html":"https://pith.science/pith/TJZ4X5Z3IOQ6GU5FOPQ3XEQ5RN","download_json":"https://pith.science/pith/TJZ4X5Z3IOQ6GU5FOPQ3XEQ5RN.json","view_paper":"https://pith.science/paper/TJZ4X5Z3","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2006.04768&json=true","fetch_graph":"https://pith.science/api/pith-number/TJZ4X5Z3IOQ6GU5FOPQ3XEQ5RN/graph.json","fetch_events":"https://pith.science/api/pith-number/TJZ4X5Z3IOQ6GU5FOPQ3XEQ5RN/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/TJZ4X5Z3IOQ6GU5FOPQ3XEQ5RN/action/timestamp_anchor","attest_storage":"https://pith.science/pith/TJZ4X5Z3IOQ6GU5FOPQ3XEQ5RN/action/storage_attestation","attest_author":"https://pith.science/pith/TJZ4X5Z3IOQ6GU5FOPQ3XEQ5RN/action/author_attestation","sign_citation":"https://pith.science/pith/TJZ4X5Z3IOQ6GU5FOPQ3XEQ5RN/action/citation_signature","submit_replication":"https://pith.science/pith/TJZ4X5Z3IOQ6GU5FOPQ3XEQ5RN/action/replication_record"}},"created_at":"2026-07-05T01:09:59.490931+00:00","updated_at":"2026-07-05T01:09:59.490931+00:00"}