{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:BZLA55XC5MCMQFCD645SAWIMJR","short_pith_number":"pith:BZLA55XC","schema_version":"1.0","canonical_sha256":"0e560ef6e2eb04c81443f73b20590c4c7b842c126b41bbf19bdad1a2f49858e1","source":{"kind":"arxiv","id":"2508.17734","version":1},"attestation_state":"computed","paper":{"title":"Layerwise Importance Analysis of Feed-Forward Networks in Transformer-based Language Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Jaesung Lee, Jun Suzuki, Kazuki Yano, Keigo Shibata, Ryosuke Takahashi, Wataru Ikeda","submitted_at":"2025-08-25T07:21:49Z","abstract_excerpt":"This study investigates the layerwise importance of feed-forward networks (FFNs) in Transformer-based language models during pretraining. We introduce an experimental approach that, while maintaining the total parameter count, increases the FFN dimensions in some layers and completely removes the FFNs from other layers. Furthermore, since our focus is on the importance of FFNs during pretraining, we train models from scratch to examine whether the importance of FFNs varies depending on their layer positions, rather than using publicly available pretrained models as is frequently done. Through "},"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":"2508.17734","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2025-08-25T07:21:49Z","cross_cats_sorted":[],"title_canon_sha256":"33f135b2bce5172f66b3371d4c4565c35774f28e5bb8227f00414dea8e607828","abstract_canon_sha256":"0010d56d76cde161c73e24030a8198b9b8343ecb348b17848b0d0cebc4fbffe4"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:58:46.746534Z","signature_b64":"OTUbHyBuzE6d4KZ13zCqd6I5mbMDVG95+SNgzKjkY+VVRFTAuEyj4MhBLkRLaGHWXc0u032S9ZeN+waKFv+/Aw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"0e560ef6e2eb04c81443f73b20590c4c7b842c126b41bbf19bdad1a2f49858e1","last_reissued_at":"2026-07-05T11:58:46.745967Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:58:46.745967Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Layerwise Importance Analysis of Feed-Forward Networks in Transformer-based Language Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Jaesung Lee, Jun Suzuki, Kazuki Yano, Keigo Shibata, Ryosuke Takahashi, Wataru Ikeda","submitted_at":"2025-08-25T07:21:49Z","abstract_excerpt":"This study investigates the layerwise importance of feed-forward networks (FFNs) in Transformer-based language models during pretraining. We introduce an experimental approach that, while maintaining the total parameter count, increases the FFN dimensions in some layers and completely removes the FFNs from other layers. Furthermore, since our focus is on the importance of FFNs during pretraining, we train models from scratch to examine whether the importance of FFNs varies depending on their layer positions, rather than using publicly available pretrained models as is frequently done. Through "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2508.17734","kind":"arxiv","version":1},"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/2508.17734/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":"2508.17734","created_at":"2026-07-05T11:58:46.746045+00:00"},{"alias_kind":"arxiv_version","alias_value":"2508.17734v1","created_at":"2026-07-05T11:58:46.746045+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2508.17734","created_at":"2026-07-05T11:58:46.746045+00:00"},{"alias_kind":"pith_short_12","alias_value":"BZLA55XC5MCM","created_at":"2026-07-05T11:58:46.746045+00:00"},{"alias_kind":"pith_short_16","alias_value":"BZLA55XC5MCMQFCD","created_at":"2026-07-05T11:58:46.746045+00:00"},{"alias_kind":"pith_short_8","alias_value":"BZLA55XC","created_at":"2026-07-05T11:58:46.746045+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":3,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.26620","citing_title":"Discovering Millions of Interpretable Features with Sparse Autoencoders","ref_index":26,"is_internal_anchor":false},{"citing_arxiv_id":"2606.23670","citing_title":"Tapered Language Models","ref_index":16,"is_internal_anchor":false},{"citing_arxiv_id":"2606.18246","citing_title":"Variable-Width Transformers","ref_index":17,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/BZLA55XC5MCMQFCD645SAWIMJR","json":"https://pith.science/pith/BZLA55XC5MCMQFCD645SAWIMJR.json","graph_json":"https://pith.science/api/pith-number/BZLA55XC5MCMQFCD645SAWIMJR/graph.json","events_json":"https://pith.science/api/pith-number/BZLA55XC5MCMQFCD645SAWIMJR/events.json","paper":"https://pith.science/paper/BZLA55XC"},"agent_actions":{"view_html":"https://pith.science/pith/BZLA55XC5MCMQFCD645SAWIMJR","download_json":"https://pith.science/pith/BZLA55XC5MCMQFCD645SAWIMJR.json","view_paper":"https://pith.science/paper/BZLA55XC","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2508.17734&json=true","fetch_graph":"https://pith.science/api/pith-number/BZLA55XC5MCMQFCD645SAWIMJR/graph.json","fetch_events":"https://pith.science/api/pith-number/BZLA55XC5MCMQFCD645SAWIMJR/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/BZLA55XC5MCMQFCD645SAWIMJR/action/timestamp_anchor","attest_storage":"https://pith.science/pith/BZLA55XC5MCMQFCD645SAWIMJR/action/storage_attestation","attest_author":"https://pith.science/pith/BZLA55XC5MCMQFCD645SAWIMJR/action/author_attestation","sign_citation":"https://pith.science/pith/BZLA55XC5MCMQFCD645SAWIMJR/action/citation_signature","submit_replication":"https://pith.science/pith/BZLA55XC5MCMQFCD645SAWIMJR/action/replication_record"}},"created_at":"2026-07-05T11:58:46.746045+00:00","updated_at":"2026-07-05T11:58:46.746045+00:00"}