{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2021:BWMG7LGANUKZPIPHBE6ITENRNE","short_pith_number":"pith:BWMG7LGA","schema_version":"1.0","canonical_sha256":"0d986facc06d1597a1e7093c8991b1693eaa3a6d17d45c8e8272795cfbbfd855","source":{"kind":"arxiv","id":"2106.08181","version":2},"attestation_state":"computed","paper":{"title":"Direction is what you need: Improving Word Embedding Compression in Large Language Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Jacek Tabor, Karl Aberer, Klaudia Ba{\\l}azy, Mohammadreza Banaei, R\\'emi Lebret","submitted_at":"2021-06-15T14:28:00Z","abstract_excerpt":"The adoption of Transformer-based models in natural language processing (NLP) has led to great success using a massive number of parameters. However, due to deployment constraints in edge devices, there has been a rising interest in the compression of these models to improve their inference time and memory footprint. This paper presents a novel loss objective to compress token embeddings in the Transformer-based models by leveraging an AutoEncoder architecture. More specifically, we emphasize the importance of the direction of compressed embeddings with respect to original uncompressed embeddi"},"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":"2106.08181","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2021-06-15T14:28:00Z","cross_cats_sorted":[],"title_canon_sha256":"39bb543a2238932196aeda94b4d288353a5f2cf265a4d1f5eded155fcc9a9ee6","abstract_canon_sha256":"4caf4848c36def707ee26142305df63b6375ce80e59f811accd78d30b83ba761"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T03:02:55.767353Z","signature_b64":"ORtfKJamvbpc+hkYgA3+1S6PrIJCcySxNKQwpYAByE2BpPM5R8zlPT3Boh5qkQUNBNzqrtb7NWOkf5H27zvrAw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"0d986facc06d1597a1e7093c8991b1693eaa3a6d17d45c8e8272795cfbbfd855","last_reissued_at":"2026-07-05T03:02:55.766812Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T03:02:55.766812Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Direction is what you need: Improving Word Embedding Compression in Large Language Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Jacek Tabor, Karl Aberer, Klaudia Ba{\\l}azy, Mohammadreza Banaei, R\\'emi Lebret","submitted_at":"2021-06-15T14:28:00Z","abstract_excerpt":"The adoption of Transformer-based models in natural language processing (NLP) has led to great success using a massive number of parameters. However, due to deployment constraints in edge devices, there has been a rising interest in the compression of these models to improve their inference time and memory footprint. This paper presents a novel loss objective to compress token embeddings in the Transformer-based models by leveraging an AutoEncoder architecture. More specifically, we emphasize the importance of the direction of compressed embeddings with respect to original uncompressed embeddi"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2106.08181","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/2106.08181/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":"2106.08181","created_at":"2026-07-05T03:02:55.766877+00:00"},{"alias_kind":"arxiv_version","alias_value":"2106.08181v2","created_at":"2026-07-05T03:02:55.766877+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2106.08181","created_at":"2026-07-05T03:02:55.766877+00:00"},{"alias_kind":"pith_short_12","alias_value":"BWMG7LGANUKZ","created_at":"2026-07-05T03:02:55.766877+00:00"},{"alias_kind":"pith_short_16","alias_value":"BWMG7LGANUKZPIPH","created_at":"2026-07-05T03:02:55.766877+00:00"},{"alias_kind":"pith_short_8","alias_value":"BWMG7LGA","created_at":"2026-07-05T03:02:55.766877+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2506.13514","citing_title":"TensorSLM: Energy-efficient Embedding Compression of Sub-billion Parameter Language Models on Low-end Devices","ref_index":5,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/BWMG7LGANUKZPIPHBE6ITENRNE","json":"https://pith.science/pith/BWMG7LGANUKZPIPHBE6ITENRNE.json","graph_json":"https://pith.science/api/pith-number/BWMG7LGANUKZPIPHBE6ITENRNE/graph.json","events_json":"https://pith.science/api/pith-number/BWMG7LGANUKZPIPHBE6ITENRNE/events.json","paper":"https://pith.science/paper/BWMG7LGA"},"agent_actions":{"view_html":"https://pith.science/pith/BWMG7LGANUKZPIPHBE6ITENRNE","download_json":"https://pith.science/pith/BWMG7LGANUKZPIPHBE6ITENRNE.json","view_paper":"https://pith.science/paper/BWMG7LGA","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2106.08181&json=true","fetch_graph":"https://pith.science/api/pith-number/BWMG7LGANUKZPIPHBE6ITENRNE/graph.json","fetch_events":"https://pith.science/api/pith-number/BWMG7LGANUKZPIPHBE6ITENRNE/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/BWMG7LGANUKZPIPHBE6ITENRNE/action/timestamp_anchor","attest_storage":"https://pith.science/pith/BWMG7LGANUKZPIPHBE6ITENRNE/action/storage_attestation","attest_author":"https://pith.science/pith/BWMG7LGANUKZPIPHBE6ITENRNE/action/author_attestation","sign_citation":"https://pith.science/pith/BWMG7LGANUKZPIPHBE6ITENRNE/action/citation_signature","submit_replication":"https://pith.science/pith/BWMG7LGANUKZPIPHBE6ITENRNE/action/replication_record"}},"created_at":"2026-07-05T03:02:55.766877+00:00","updated_at":"2026-07-05T03:02:55.766877+00:00"}