{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:LLBKMI6JACZLJ3H4QGC4XPFXCY","short_pith_number":"pith:LLBKMI6J","schema_version":"1.0","canonical_sha256":"5ac2a623c900b2b4ecfc8185cbbcb7160c9eabdaf1a9096169e229527bfe6fbe","source":{"kind":"arxiv","id":"2506.18020","version":3},"attestation_state":"computed","paper":{"title":"Tight Stability Bounds for Robust Distributed Learning: Byzantine Failures Hurt Generalization More than Data Poisoning","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CR","stat.ML"],"primary_cat":"cs.LG","authors_text":"Aur\\'elien Bellet, Batiste Le Bars, Nirupam Gupta, Thomas Boudou","submitted_at":"2025-06-22T12:59:15Z","abstract_excerpt":"Robust distributed learning algorithms aim to maintain reliable performance despite the presence of misbehaving workers. Such misbehaviors are commonly modeled as \\textit{Byzantine failures}, allowing arbitrarily corrupted communication, or as \\textit{data poisoning}, a weaker form of corruption restricted to local training data. While prior work shows similar optimization guarantees for both models, an important question remains: \\textit{How do these threat models impact generalization?} We show, for the first time, a fundamental gap in generalization guarantees between the two threat models:"},"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":"2506.18020","kind":"arxiv","version":3},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-06-22T12:59:15Z","cross_cats_sorted":["cs.CR","stat.ML"],"title_canon_sha256":"09c7b4146b0a238b4af7ce661ba15e829f81d2c4d9af73da5e50d2b3585b5006","abstract_canon_sha256":"2ba540115e840014b2d46320f32f89622befa53ffd7861530cbc9722f923c12f"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-07T01:15:55.377207Z","signature_b64":"PnZ2665HCoaIVxYnYHlsxLhqpShtp6IvhOM2B5jKP3fEfA+bTVrvX+tEwgtofimYTuLqs5qJGdEUs/QWVTZ7Bw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"5ac2a623c900b2b4ecfc8185cbbcb7160c9eabdaf1a9096169e229527bfe6fbe","last_reissued_at":"2026-07-07T01:15:55.376423Z","signature_status":"signed_v1","first_computed_at":"2026-07-07T01:15:55.376423Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Tight Stability Bounds for Robust Distributed Learning: Byzantine Failures Hurt Generalization More than Data Poisoning","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CR","stat.ML"],"primary_cat":"cs.LG","authors_text":"Aur\\'elien Bellet, Batiste Le Bars, Nirupam Gupta, Thomas Boudou","submitted_at":"2025-06-22T12:59:15Z","abstract_excerpt":"Robust distributed learning algorithms aim to maintain reliable performance despite the presence of misbehaving workers. Such misbehaviors are commonly modeled as \\textit{Byzantine failures}, allowing arbitrarily corrupted communication, or as \\textit{data poisoning}, a weaker form of corruption restricted to local training data. While prior work shows similar optimization guarantees for both models, an important question remains: \\textit{How do these threat models impact generalization?} We show, for the first time, a fundamental gap in generalization guarantees between the two threat models:"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2506.18020","kind":"arxiv","version":3},"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/2506.18020/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":"2506.18020","created_at":"2026-07-07T01:15:55.376513+00:00"},{"alias_kind":"arxiv_version","alias_value":"2506.18020v3","created_at":"2026-07-07T01:15:55.376513+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2506.18020","created_at":"2026-07-07T01:15:55.376513+00:00"},{"alias_kind":"pith_short_12","alias_value":"LLBKMI6JACZL","created_at":"2026-07-07T01:15:55.376513+00:00"},{"alias_kind":"pith_short_16","alias_value":"LLBKMI6JACZLJ3H4","created_at":"2026-07-07T01:15:55.376513+00:00"},{"alias_kind":"pith_short_8","alias_value":"LLBKMI6J","created_at":"2026-07-07T01:15:55.376513+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":2,"internal_anchor_count":2,"sample":[{"citing_arxiv_id":"2607.01492","citing_title":"Unveiling the Non-Monotonic Effect of Privacy on Generalization under Byzantine Robustness","ref_index":7,"is_internal_anchor":true},{"citing_arxiv_id":"2606.28123","citing_title":"Dangerous Liaisons of Convex Learning and Non-Affine Aggregation","ref_index":1,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/LLBKMI6JACZLJ3H4QGC4XPFXCY","json":"https://pith.science/pith/LLBKMI6JACZLJ3H4QGC4XPFXCY.json","graph_json":"https://pith.science/api/pith-number/LLBKMI6JACZLJ3H4QGC4XPFXCY/graph.json","events_json":"https://pith.science/api/pith-number/LLBKMI6JACZLJ3H4QGC4XPFXCY/events.json","paper":"https://pith.science/paper/LLBKMI6J"},"agent_actions":{"view_html":"https://pith.science/pith/LLBKMI6JACZLJ3H4QGC4XPFXCY","download_json":"https://pith.science/pith/LLBKMI6JACZLJ3H4QGC4XPFXCY.json","view_paper":"https://pith.science/paper/LLBKMI6J","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2506.18020&json=true","fetch_graph":"https://pith.science/api/pith-number/LLBKMI6JACZLJ3H4QGC4XPFXCY/graph.json","fetch_events":"https://pith.science/api/pith-number/LLBKMI6JACZLJ3H4QGC4XPFXCY/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/LLBKMI6JACZLJ3H4QGC4XPFXCY/action/timestamp_anchor","attest_storage":"https://pith.science/pith/LLBKMI6JACZLJ3H4QGC4XPFXCY/action/storage_attestation","attest_author":"https://pith.science/pith/LLBKMI6JACZLJ3H4QGC4XPFXCY/action/author_attestation","sign_citation":"https://pith.science/pith/LLBKMI6JACZLJ3H4QGC4XPFXCY/action/citation_signature","submit_replication":"https://pith.science/pith/LLBKMI6JACZLJ3H4QGC4XPFXCY/action/replication_record"}},"created_at":"2026-07-07T01:15:55.376513+00:00","updated_at":"2026-07-07T01:15:55.376513+00:00"}