{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:PZYKXYCON6AJUU6FXDGT7X4WNN","short_pith_number":"pith:PZYKXYCO","schema_version":"1.0","canonical_sha256":"7e70abe04e6f809a53c5b8cd3fdf966b653c3dff5fbcd9126b66f85393e5e583","source":{"kind":"arxiv","id":"2506.21833","version":1},"attestation_state":"computed","paper":{"title":"The Cost of Avoiding Backpropagation","license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Hui Guan, Kunjal Panchal, Sunav Choudhary, Yuriy Brun","submitted_at":"2025-06-27T00:47:03Z","abstract_excerpt":"Forward-mode automatic differentiation (FmAD) and zero-order (ZO) optimization have been proposed as memory-efficient alternatives to backpropagation (BP) for gradient computation, especially in low-resource settings. However, their practical benefits remain unclear due to two key gaps: a lack of comparison against memory-efficient BP variants, such as activation checkpointing, and a lack of a unified theoretical analysis. This work presents a comprehensive theoretical and empirical comparison of BP, FmAD, and ZO methods. Our theoretical analysis shows that while FmAD, and ZO can reduce memory"},"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":"2506.21833","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","primary_cat":"cs.LG","submitted_at":"2025-06-27T00:47:03Z","cross_cats_sorted":[],"title_canon_sha256":"2d8ba3c25b29b3b4f590cb500da2651cf2c1cac79cf9549584c94091d9c5ed50","abstract_canon_sha256":"605870f2faadb01538af2e63cf2d18cfec91af29cb777c5b64ee33115aafe5a8"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:28:02.021923Z","signature_b64":"NkD4m7WMtl2Tt/NAAxFQQK2nmfeKJ6sqBpR4QubV0BDsRNWcPmnhi3hncN1m0xr7jTU5xa3dOefy4J909PWcCw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"7e70abe04e6f809a53c5b8cd3fdf966b653c3dff5fbcd9126b66f85393e5e583","last_reissued_at":"2026-07-05T11:28:02.021408Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:28:02.021408Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"The Cost of Avoiding Backpropagation","license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Hui Guan, Kunjal Panchal, Sunav Choudhary, Yuriy Brun","submitted_at":"2025-06-27T00:47:03Z","abstract_excerpt":"Forward-mode automatic differentiation (FmAD) and zero-order (ZO) optimization have been proposed as memory-efficient alternatives to backpropagation (BP) for gradient computation, especially in low-resource settings. However, their practical benefits remain unclear due to two key gaps: a lack of comparison against memory-efficient BP variants, such as activation checkpointing, and a lack of a unified theoretical analysis. This work presents a comprehensive theoretical and empirical comparison of BP, FmAD, and ZO methods. Our theoretical analysis shows that while FmAD, and ZO can reduce memory"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2506.21833","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/2506.21833/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":"2506.21833","created_at":"2026-07-05T11:28:02.021474+00:00"},{"alias_kind":"arxiv_version","alias_value":"2506.21833v1","created_at":"2026-07-05T11:28:02.021474+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2506.21833","created_at":"2026-07-05T11:28:02.021474+00:00"},{"alias_kind":"pith_short_12","alias_value":"PZYKXYCON6AJ","created_at":"2026-07-05T11:28:02.021474+00:00"},{"alias_kind":"pith_short_16","alias_value":"PZYKXYCON6AJUU6F","created_at":"2026-07-05T11:28:02.021474+00:00"},{"alias_kind":"pith_short_8","alias_value":"PZYKXYCO","created_at":"2026-07-05T11:28:02.021474+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":2,"internal_anchor_count":2,"sample":[{"citing_arxiv_id":"2606.09734","citing_title":"Adaptive directional gradients for parameterised quantum circuits","ref_index":61,"is_internal_anchor":true},{"citing_arxiv_id":"2511.11362","citing_title":"On-Device Fine-Tuning via Backprop-Free Zeroth-Order Optimization","ref_index":13,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/PZYKXYCON6AJUU6FXDGT7X4WNN","json":"https://pith.science/pith/PZYKXYCON6AJUU6FXDGT7X4WNN.json","graph_json":"https://pith.science/api/pith-number/PZYKXYCON6AJUU6FXDGT7X4WNN/graph.json","events_json":"https://pith.science/api/pith-number/PZYKXYCON6AJUU6FXDGT7X4WNN/events.json","paper":"https://pith.science/paper/PZYKXYCO"},"agent_actions":{"view_html":"https://pith.science/pith/PZYKXYCON6AJUU6FXDGT7X4WNN","download_json":"https://pith.science/pith/PZYKXYCON6AJUU6FXDGT7X4WNN.json","view_paper":"https://pith.science/paper/PZYKXYCO","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2506.21833&json=true","fetch_graph":"https://pith.science/api/pith-number/PZYKXYCON6AJUU6FXDGT7X4WNN/graph.json","fetch_events":"https://pith.science/api/pith-number/PZYKXYCON6AJUU6FXDGT7X4WNN/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/PZYKXYCON6AJUU6FXDGT7X4WNN/action/timestamp_anchor","attest_storage":"https://pith.science/pith/PZYKXYCON6AJUU6FXDGT7X4WNN/action/storage_attestation","attest_author":"https://pith.science/pith/PZYKXYCON6AJUU6FXDGT7X4WNN/action/author_attestation","sign_citation":"https://pith.science/pith/PZYKXYCON6AJUU6FXDGT7X4WNN/action/citation_signature","submit_replication":"https://pith.science/pith/PZYKXYCON6AJUU6FXDGT7X4WNN/action/replication_record"}},"created_at":"2026-07-05T11:28:02.021474+00:00","updated_at":"2026-07-05T11:28:02.021474+00:00"}