{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2023:LJ7RWJ2MFDSZB6D3DIY36SSAGQ","short_pith_number":"pith:LJ7RWJ2M","schema_version":"1.0","canonical_sha256":"5a7f1b274c28e590f87b1a31bf4a4034184fd1f63057914630567b4196e307bf","source":{"kind":"arxiv","id":"2306.00008","version":2},"attestation_state":"computed","paper":{"title":"Brainformers: Trading Simplicity for Efficiency","license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.LG","authors_text":"Andrew Dai, Chang Lan, Claire Cui, Da Huang, Daiyi Peng, David So, James Laudon, Jeff Dean, Nan Du, Quoc Le, Siamak Shakeri, Yanping Huang, Yanqi Zhou, Yifeng Lu, Zhifeng Chen","submitted_at":"2023-05-29T18:42:01Z","abstract_excerpt":"Transformers are central to recent successes in natural language processing and computer vision. Transformers have a mostly uniform backbone where layers alternate between feed-forward and self-attention in order to build a deep network. Here we investigate this design choice and find that more complex blocks that have different permutations of layer primitives can be more efficient. Using this insight, we develop a complex block, named Brainformer, that consists of a diverse sets of layers such as sparsely gated feed-forward layers, dense feed-forward layers, attention layers, and various for"},"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":"2306.00008","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","primary_cat":"cs.LG","submitted_at":"2023-05-29T18:42:01Z","cross_cats_sorted":["cs.CL"],"title_canon_sha256":"785d109acf1d3053fa787413cc65c12fa2056f47481b7c0438fa1af9cebdbea5","abstract_canon_sha256":"46b2ceba45918f9ac4afd19d6883f6e48cff52467466b715511237ce130ef93c"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T08:11:53.079932Z","signature_b64":"I91Ta1O+pPsM3BZVTsg61oZVukatwwEqTfvKzpHXkUdSbvPPNZIR9l/MSxkos2yYU/3ff2nDVLJbfV98j/wSAg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"5a7f1b274c28e590f87b1a31bf4a4034184fd1f63057914630567b4196e307bf","last_reissued_at":"2026-07-05T08:11:53.079515Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T08:11:53.079515Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Brainformers: Trading Simplicity for Efficiency","license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.LG","authors_text":"Andrew Dai, Chang Lan, Claire Cui, Da Huang, Daiyi Peng, David So, James Laudon, Jeff Dean, Nan Du, Quoc Le, Siamak Shakeri, Yanping Huang, Yanqi Zhou, Yifeng Lu, Zhifeng Chen","submitted_at":"2023-05-29T18:42:01Z","abstract_excerpt":"Transformers are central to recent successes in natural language processing and computer vision. Transformers have a mostly uniform backbone where layers alternate between feed-forward and self-attention in order to build a deep network. Here we investigate this design choice and find that more complex blocks that have different permutations of layer primitives can be more efficient. Using this insight, we develop a complex block, named Brainformer, that consists of a diverse sets of layers such as sparsely gated feed-forward layers, dense feed-forward layers, attention layers, and various for"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2306.00008","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/2306.00008/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":"2306.00008","created_at":"2026-07-05T08:11:53.079578+00:00"},{"alias_kind":"arxiv_version","alias_value":"2306.00008v2","created_at":"2026-07-05T08:11:53.079578+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2306.00008","created_at":"2026-07-05T08:11:53.079578+00:00"},{"alias_kind":"pith_short_12","alias_value":"LJ7RWJ2MFDSZ","created_at":"2026-07-05T08:11:53.079578+00:00"},{"alias_kind":"pith_short_16","alias_value":"LJ7RWJ2MFDSZB6D3","created_at":"2026-07-05T08:11:53.079578+00:00"},{"alias_kind":"pith_short_8","alias_value":"LJ7RWJ2M","created_at":"2026-07-05T08:11:53.079578+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2509.06518","citing_title":"Crown, Frame, Reverse: Layer-Wise Scaling Variants for LLM Pre-Training","ref_index":31,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/LJ7RWJ2MFDSZB6D3DIY36SSAGQ","json":"https://pith.science/pith/LJ7RWJ2MFDSZB6D3DIY36SSAGQ.json","graph_json":"https://pith.science/api/pith-number/LJ7RWJ2MFDSZB6D3DIY36SSAGQ/graph.json","events_json":"https://pith.science/api/pith-number/LJ7RWJ2MFDSZB6D3DIY36SSAGQ/events.json","paper":"https://pith.science/paper/LJ7RWJ2M"},"agent_actions":{"view_html":"https://pith.science/pith/LJ7RWJ2MFDSZB6D3DIY36SSAGQ","download_json":"https://pith.science/pith/LJ7RWJ2MFDSZB6D3DIY36SSAGQ.json","view_paper":"https://pith.science/paper/LJ7RWJ2M","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2306.00008&json=true","fetch_graph":"https://pith.science/api/pith-number/LJ7RWJ2MFDSZB6D3DIY36SSAGQ/graph.json","fetch_events":"https://pith.science/api/pith-number/LJ7RWJ2MFDSZB6D3DIY36SSAGQ/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/LJ7RWJ2MFDSZB6D3DIY36SSAGQ/action/timestamp_anchor","attest_storage":"https://pith.science/pith/LJ7RWJ2MFDSZB6D3DIY36SSAGQ/action/storage_attestation","attest_author":"https://pith.science/pith/LJ7RWJ2MFDSZB6D3DIY36SSAGQ/action/author_attestation","sign_citation":"https://pith.science/pith/LJ7RWJ2MFDSZB6D3DIY36SSAGQ/action/citation_signature","submit_replication":"https://pith.science/pith/LJ7RWJ2MFDSZB6D3DIY36SSAGQ/action/replication_record"}},"created_at":"2026-07-05T08:11:53.079578+00:00","updated_at":"2026-07-05T08:11:53.079578+00:00"}