{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2022:T3BM574GUCDVWWQ4OLXBNPN5FM","short_pith_number":"pith:T3BM574G","schema_version":"1.0","canonical_sha256":"9ec2ceff86a0875b5a1c72ee16bdbd2b2b3a9673d5bcec115e0793d155feb151","source":{"kind":"arxiv","id":"2203.02094","version":2},"attestation_state":"computed","paper":{"title":"LiteTransformerSearch: Training-free Neural Architecture Search for Efficient Language Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.LG","authors_text":"Caio C. T. Mendes, Debadeepta Dey, Farinaz Koushanfar, Gustavo H. de Rosa, Mojan Javaheripi, Sebastien Bubeck, Shital Shah, Subhabrata Mukherjee, Tomasz L. Religa","submitted_at":"2022-03-04T02:10:43Z","abstract_excerpt":"The Transformer architecture is ubiquitously used as the building block of large-scale autoregressive language models. However, finding architectures with the optimal trade-off between task performance (perplexity) and hardware constraints like peak memory utilization and latency is non-trivial. This is exacerbated by the proliferation of various hardware. We leverage the somewhat surprising empirical observation that the number of decoder parameters in autoregressive Transformers has a high rank correlation with task performance, irrespective of the architecture topology. This observation org"},"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":"2203.02094","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2022-03-04T02:10:43Z","cross_cats_sorted":["cs.CL"],"title_canon_sha256":"1f414d1f293c1e891b61e4850b6cf3b6d6a096ee9b7d69f97d513e8223dd09bc","abstract_canon_sha256":"f32b8e0f77179d7160064de78585be9848f4ad5cfd18d635385ad8731c98c1c0"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T05:07:21.872214Z","signature_b64":"JRlKAAtPPNfMwqxONRbKe/7DttshI9HWQQ0zly+f47wRSVDn/sdtwpuH5iW+g8c8BLpSMEnxjN8QD+FoiRIBBA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"9ec2ceff86a0875b5a1c72ee16bdbd2b2b3a9673d5bcec115e0793d155feb151","last_reissued_at":"2026-07-05T05:07:21.871768Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T05:07:21.871768Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"LiteTransformerSearch: Training-free Neural Architecture Search for Efficient Language Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.LG","authors_text":"Caio C. T. Mendes, Debadeepta Dey, Farinaz Koushanfar, Gustavo H. de Rosa, Mojan Javaheripi, Sebastien Bubeck, Shital Shah, Subhabrata Mukherjee, Tomasz L. Religa","submitted_at":"2022-03-04T02:10:43Z","abstract_excerpt":"The Transformer architecture is ubiquitously used as the building block of large-scale autoregressive language models. However, finding architectures with the optimal trade-off between task performance (perplexity) and hardware constraints like peak memory utilization and latency is non-trivial. This is exacerbated by the proliferation of various hardware. We leverage the somewhat surprising empirical observation that the number of decoder parameters in autoregressive Transformers has a high rank correlation with task performance, irrespective of the architecture topology. This observation org"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2203.02094","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/2203.02094/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":"2203.02094","created_at":"2026-07-05T05:07:21.871818+00:00"},{"alias_kind":"arxiv_version","alias_value":"2203.02094v2","created_at":"2026-07-05T05:07:21.871818+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2203.02094","created_at":"2026-07-05T05:07:21.871818+00:00"},{"alias_kind":"pith_short_12","alias_value":"T3BM574GUCDV","created_at":"2026-07-05T05:07:21.871818+00:00"},{"alias_kind":"pith_short_16","alias_value":"T3BM574GUCDVWWQ4","created_at":"2026-07-05T05:07:21.871818+00:00"},{"alias_kind":"pith_short_8","alias_value":"T3BM574G","created_at":"2026-07-05T05:07:21.871818+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2402.13228","citing_title":"Smaug: Fixing Failure Modes of Preference Optimisation with DPO-Positive","ref_index":198,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/T3BM574GUCDVWWQ4OLXBNPN5FM","json":"https://pith.science/pith/T3BM574GUCDVWWQ4OLXBNPN5FM.json","graph_json":"https://pith.science/api/pith-number/T3BM574GUCDVWWQ4OLXBNPN5FM/graph.json","events_json":"https://pith.science/api/pith-number/T3BM574GUCDVWWQ4OLXBNPN5FM/events.json","paper":"https://pith.science/paper/T3BM574G"},"agent_actions":{"view_html":"https://pith.science/pith/T3BM574GUCDVWWQ4OLXBNPN5FM","download_json":"https://pith.science/pith/T3BM574GUCDVWWQ4OLXBNPN5FM.json","view_paper":"https://pith.science/paper/T3BM574G","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2203.02094&json=true","fetch_graph":"https://pith.science/api/pith-number/T3BM574GUCDVWWQ4OLXBNPN5FM/graph.json","fetch_events":"https://pith.science/api/pith-number/T3BM574GUCDVWWQ4OLXBNPN5FM/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/T3BM574GUCDVWWQ4OLXBNPN5FM/action/timestamp_anchor","attest_storage":"https://pith.science/pith/T3BM574GUCDVWWQ4OLXBNPN5FM/action/storage_attestation","attest_author":"https://pith.science/pith/T3BM574GUCDVWWQ4OLXBNPN5FM/action/author_attestation","sign_citation":"https://pith.science/pith/T3BM574GUCDVWWQ4OLXBNPN5FM/action/citation_signature","submit_replication":"https://pith.science/pith/T3BM574GUCDVWWQ4OLXBNPN5FM/action/replication_record"}},"created_at":"2026-07-05T05:07:21.871818+00:00","updated_at":"2026-07-05T05:07:21.871818+00:00"}