{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2023:J4GNPTF2RLK373NDUMOEMINFAA","short_pith_number":"pith:J4GNPTF2","canonical_record":{"source":{"id":"2303.00783","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2023-03-01T19:10:05Z","cross_cats_sorted":["cs.CR","cs.NE","stat.ML"],"title_canon_sha256":"da0def6f581d5846bf92526502ce76449b7b11eb033593d077235eb02fcaee34","abstract_canon_sha256":"cb250f8a42160fd72dc61416330dc2a4044a7ab9dec44d42adabb865032648df"},"schema_version":"1.0"},"canonical_sha256":"4f0cd7ccba8ad5bfeda3a31c4621a5000e5e0a457fbb53bc40d1ff16a7454bf8","source":{"kind":"arxiv","id":"2303.00783","version":2},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2303.00783","created_at":"2026-07-05T07:13:21Z"},{"alias_kind":"arxiv_version","alias_value":"2303.00783v2","created_at":"2026-07-05T07:13:21Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2303.00783","created_at":"2026-07-05T07:13:21Z"},{"alias_kind":"pith_short_12","alias_value":"J4GNPTF2RLK3","created_at":"2026-07-05T07:13:21Z"},{"alias_kind":"pith_short_16","alias_value":"J4GNPTF2RLK373ND","created_at":"2026-07-05T07:13:21Z"},{"alias_kind":"pith_short_8","alias_value":"J4GNPTF2","created_at":"2026-07-05T07:13:21Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2023:J4GNPTF2RLK373NDUMOEMINFAA","target":"record","payload":{"canonical_record":{"source":{"id":"2303.00783","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2023-03-01T19:10:05Z","cross_cats_sorted":["cs.CR","cs.NE","stat.ML"],"title_canon_sha256":"da0def6f581d5846bf92526502ce76449b7b11eb033593d077235eb02fcaee34","abstract_canon_sha256":"cb250f8a42160fd72dc61416330dc2a4044a7ab9dec44d42adabb865032648df"},"schema_version":"1.0"},"canonical_sha256":"4f0cd7ccba8ad5bfeda3a31c4621a5000e5e0a457fbb53bc40d1ff16a7454bf8","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T07:13:21.627085Z","signature_b64":"KENPoyHA/bXYyOEbxCFTl8Ov63jdCjmYWLRpKoU3KQymV24y0KfxmmYjL9Xa34xCzn97z7FfV6ZnTa0bJieYDg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"4f0cd7ccba8ad5bfeda3a31c4621a5000e5e0a457fbb53bc40d1ff16a7454bf8","last_reissued_at":"2026-07-05T07:13:21.626576Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T07:13:21.626576Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2303.00783","source_version":2,"attestation_state":"computed"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-07-05T07:13:21Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"1p3hkTGir8YIMcM/CFii121i0ir9CCM+lu28j68KO6Dwjre3MZMeS0nRJk+jPUJE0bptn3XVkxitLAomLE5mBg==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-07-07T06:39:43.984693Z"},"content_sha256":"3ae766085bf4fc238e3a49a2e34b207195590289d3b34fa49189a857345b22a4","schema_version":"1.0","event_id":"sha256:3ae766085bf4fc238e3a49a2e34b207195590289d3b34fa49189a857345b22a4"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2023:J4GNPTF2RLK373NDUMOEMINFAA","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"Adversarial Examples Exist in Two-Layer ReLU Networks for Low Dimensional Linear Subspaces","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CR","cs.NE","stat.ML"],"primary_cat":"cs.LG","authors_text":"Gal Vardi, Gilad Yehudai, Odelia Melamed","submitted_at":"2023-03-01T19:10:05Z","abstract_excerpt":"Despite a great deal of research, it is still not well-understood why trained neural networks are highly vulnerable to adversarial examples. In this work we focus on two-layer neural networks trained using data which lie on a low dimensional linear subspace. We show that standard gradient methods lead to non-robust neural networks, namely, networks which have large gradients in directions orthogonal to the data subspace, and are susceptible to small adversarial $L_2$-perturbations in these directions. Moreover, we show that decreasing the initialization scale of the training algorithm, or addi"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2303.00783","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/2303.00783/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"},"verdict_id":null},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-07-05T07:13:21Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"kSqhu5FagclsfbUdbKHikwVKuaIYyeHd+3XgiOyjhH7v/J/F4NTQv3Kt0SKV1FFwiwBbzqZHuo/duXFZGRs2DQ==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-07-07T06:39:43.985073Z"},"content_sha256":"69ba490688662b7635d6690c746fe156d5f08405b49ea1708fbceee04067aa53","schema_version":"1.0","event_id":"sha256:69ba490688662b7635d6690c746fe156d5f08405b49ea1708fbceee04067aa53"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/J4GNPTF2RLK373NDUMOEMINFAA/bundle.json","state_url":"https://pith.science/pith/J4GNPTF2RLK373NDUMOEMINFAA/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/J4GNPTF2RLK373NDUMOEMINFAA/bundle.json","status":"primary"}],"public_keys":[{"key_id":"pith-v1-2026-05","algorithm":"ed25519","format":"raw","public_key_b64":"stVStoiQhXFxp4s2pdzPNoqVNBMojDU/fJ2db5S3CbM=","public_key_hex":"b2d552b68890857171a78b36a5dccf368a953413288c353f7c9d9d6f94b709b3","fingerprint_sha256_b32_first128bits":"RVFV5Z2OI2J3ZUO7ERDEBCYNKS","fingerprint_sha256_hex":"8d4b5ee74e4693bcd1df2446408b0d54","rotates_at":null,"url":"https://pith.science/pith-signing-key.json","notes":"Pith uses this Ed25519 key to sign canonical record SHA-256 digests. Verify with: ed25519_verify(public_key, message=canonical_sha256_bytes, signature=base64decode(signature_b64))."}],"merge_version":"pith-open-graph-merge-v1","built_at":"2026-07-07T06:39:43Z","links":{"resolver":"https://pith.science/pith/J4GNPTF2RLK373NDUMOEMINFAA","bundle":"https://pith.science/pith/J4GNPTF2RLK373NDUMOEMINFAA/bundle.json","state":"https://pith.science/pith/J4GNPTF2RLK373NDUMOEMINFAA/state.json","well_known_bundle":"https://pith.science/.well-known/pith/J4GNPTF2RLK373NDUMOEMINFAA/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2023:J4GNPTF2RLK373NDUMOEMINFAA","merge_version":"pith-open-graph-merge-v1","event_count":2,"valid_event_count":2,"invalid_event_count":0,"equivocation_count":0,"current":{"canonical_record":{"metadata":{"abstract_canon_sha256":"cb250f8a42160fd72dc61416330dc2a4044a7ab9dec44d42adabb865032648df","cross_cats_sorted":["cs.CR","cs.NE","stat.ML"],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2023-03-01T19:10:05Z","title_canon_sha256":"da0def6f581d5846bf92526502ce76449b7b11eb033593d077235eb02fcaee34"},"schema_version":"1.0","source":{"id":"2303.00783","kind":"arxiv","version":2}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2303.00783","created_at":"2026-07-05T07:13:21Z"},{"alias_kind":"arxiv_version","alias_value":"2303.00783v2","created_at":"2026-07-05T07:13:21Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2303.00783","created_at":"2026-07-05T07:13:21Z"},{"alias_kind":"pith_short_12","alias_value":"J4GNPTF2RLK3","created_at":"2026-07-05T07:13:21Z"},{"alias_kind":"pith_short_16","alias_value":"J4GNPTF2RLK373ND","created_at":"2026-07-05T07:13:21Z"},{"alias_kind":"pith_short_8","alias_value":"J4GNPTF2","created_at":"2026-07-05T07:13:21Z"}],"graph_snapshots":[{"event_id":"sha256:69ba490688662b7635d6690c746fe156d5f08405b49ea1708fbceee04067aa53","target":"graph","created_at":"2026-07-05T07:13:21Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"graph_snapshot":{"author_claims":{"count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","strong_count":0},"builder_version":"pith-number-builder-2026-05-17-v1","claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"integrity":{"available":true,"clean":true,"detectors_run":[],"endpoint":"/pith/2303.00783/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Despite a great deal of research, it is still not well-understood why trained neural networks are highly vulnerable to adversarial examples. In this work we focus on two-layer neural networks trained using data which lie on a low dimensional linear subspace. We show that standard gradient methods lead to non-robust neural networks, namely, networks which have large gradients in directions orthogonal to the data subspace, and are susceptible to small adversarial $L_2$-perturbations in these directions. Moreover, we show that decreasing the initialization scale of the training algorithm, or addi","authors_text":"Gal Vardi, Gilad Yehudai, Odelia Melamed","cross_cats":["cs.CR","cs.NE","stat.ML"],"headline":"","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2023-03-01T19:10:05Z","title":"Adversarial Examples Exist in Two-Layer ReLU Networks for Low Dimensional Linear Subspaces"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2303.00783","kind":"arxiv","version":2},"verdict":{"created_at":null,"id":null,"model_set":{},"one_line_summary":"","pipeline_version":null,"pith_extraction_headline":"","strongest_claim":"","weakest_assumption":""}},"verdict_id":null}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:3ae766085bf4fc238e3a49a2e34b207195590289d3b34fa49189a857345b22a4","target":"record","created_at":"2026-07-05T07:13:21Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"attestation_state":"computed","canonical_record":{"metadata":{"abstract_canon_sha256":"cb250f8a42160fd72dc61416330dc2a4044a7ab9dec44d42adabb865032648df","cross_cats_sorted":["cs.CR","cs.NE","stat.ML"],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2023-03-01T19:10:05Z","title_canon_sha256":"da0def6f581d5846bf92526502ce76449b7b11eb033593d077235eb02fcaee34"},"schema_version":"1.0","source":{"id":"2303.00783","kind":"arxiv","version":2}},"canonical_sha256":"4f0cd7ccba8ad5bfeda3a31c4621a5000e5e0a457fbb53bc40d1ff16a7454bf8","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"4f0cd7ccba8ad5bfeda3a31c4621a5000e5e0a457fbb53bc40d1ff16a7454bf8","first_computed_at":"2026-07-05T07:13:21.626576Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-07-05T07:13:21.626576Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"KENPoyHA/bXYyOEbxCFTl8Ov63jdCjmYWLRpKoU3KQymV24y0KfxmmYjL9Xa34xCzn97z7FfV6ZnTa0bJieYDg==","signature_status":"signed_v1","signed_at":"2026-07-05T07:13:21.627085Z","signed_message":"canonical_sha256_bytes"},"source_id":"2303.00783","source_kind":"arxiv","source_version":2}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:3ae766085bf4fc238e3a49a2e34b207195590289d3b34fa49189a857345b22a4","sha256:69ba490688662b7635d6690c746fe156d5f08405b49ea1708fbceee04067aa53"],"state_sha256":"a42d2b8c42dacb684431663715b25662fbe806a79082b353fe7367133691bd4c"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"SWPiO6p7pb71SeGK1UiFP9QaPdoCvuqwMjQE9Pd64ylv6qF+1JxZ6cmDCuIA6q/jjG3olPM8PBWIf1MOt4LMBw==","signed_message":"bundle_sha256_bytes","signed_at":"2026-07-07T06:39:43.987493Z","bundle_sha256":"b50ddb8632edde7a6beb9dbf9054daba4f0dc3572074dfbbe56e6e38b431a24a"}}