{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2021:KCBKT6XZ3WVGDOD3OF6UCNHA2Q","short_pith_number":"pith:KCBKT6XZ","schema_version":"1.0","canonical_sha256":"5082a9faf9ddaa61b87b717d4134e0d414712d97f927df5aba3a02194ea565e2","source":{"kind":"arxiv","id":"2108.04217","version":1},"attestation_state":"computed","paper":{"title":"ROPUST: Improving Robustness through Fine-tuning with Photonic Processors and Synthetic Gradients","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CV","authors_text":"Alessandro Cappelli, Iacopo Poli, Julien Launay, Laurent Meunier, Ruben Ohana","submitted_at":"2021-07-06T12:03:36Z","abstract_excerpt":"Robustness to adversarial attacks is typically obtained through expensive adversarial training with Projected Gradient Descent. Here we introduce ROPUST, a remarkably simple and efficient method to leverage robust pre-trained models and further increase their robustness, at no cost in natural accuracy. Our technique relies on the use of an Optical Processing Unit (OPU), a photonic co-processor, and a fine-tuning step performed with Direct Feedback Alignment, a synthetic gradient training scheme. We test our method on nine different models against four attacks in RobustBench, consistently impro"},"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":"2108.04217","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CV","submitted_at":"2021-07-06T12:03:36Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"3084a12364578613e95c6add283d7a511f6124a4dc3afe9bb4b0d1678a888c3c","abstract_canon_sha256":"c4ca5d65d38575db2748de3bc38ddd1e292a95b7e801e56a6c48907e04b08c14"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T03:04:13.889131Z","signature_b64":"QOW3IzIbgdU9kayBENh2zha92RuzEp7CU1kv18kQpfHRVC7uBxav3fnnqzog8sGypmdj6O8BOvQ4KQrFP4cQBQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"5082a9faf9ddaa61b87b717d4134e0d414712d97f927df5aba3a02194ea565e2","last_reissued_at":"2026-07-05T03:04:13.888725Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T03:04:13.888725Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"ROPUST: Improving Robustness through Fine-tuning with Photonic Processors and Synthetic Gradients","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CV","authors_text":"Alessandro Cappelli, Iacopo Poli, Julien Launay, Laurent Meunier, Ruben Ohana","submitted_at":"2021-07-06T12:03:36Z","abstract_excerpt":"Robustness to adversarial attacks is typically obtained through expensive adversarial training with Projected Gradient Descent. Here we introduce ROPUST, a remarkably simple and efficient method to leverage robust pre-trained models and further increase their robustness, at no cost in natural accuracy. Our technique relies on the use of an Optical Processing Unit (OPU), a photonic co-processor, and a fine-tuning step performed with Direct Feedback Alignment, a synthetic gradient training scheme. We test our method on nine different models against four attacks in RobustBench, consistently impro"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2108.04217","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/2108.04217/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":"2108.04217","created_at":"2026-07-05T03:04:13.888779+00:00"},{"alias_kind":"arxiv_version","alias_value":"2108.04217v1","created_at":"2026-07-05T03:04:13.888779+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2108.04217","created_at":"2026-07-05T03:04:13.888779+00:00"},{"alias_kind":"pith_short_12","alias_value":"KCBKT6XZ3WVG","created_at":"2026-07-05T03:04:13.888779+00:00"},{"alias_kind":"pith_short_16","alias_value":"KCBKT6XZ3WVGDOD3","created_at":"2026-07-05T03:04:13.888779+00:00"},{"alias_kind":"pith_short_8","alias_value":"KCBKT6XZ","created_at":"2026-07-05T03:04:13.888779+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/KCBKT6XZ3WVGDOD3OF6UCNHA2Q","json":"https://pith.science/pith/KCBKT6XZ3WVGDOD3OF6UCNHA2Q.json","graph_json":"https://pith.science/api/pith-number/KCBKT6XZ3WVGDOD3OF6UCNHA2Q/graph.json","events_json":"https://pith.science/api/pith-number/KCBKT6XZ3WVGDOD3OF6UCNHA2Q/events.json","paper":"https://pith.science/paper/KCBKT6XZ"},"agent_actions":{"view_html":"https://pith.science/pith/KCBKT6XZ3WVGDOD3OF6UCNHA2Q","download_json":"https://pith.science/pith/KCBKT6XZ3WVGDOD3OF6UCNHA2Q.json","view_paper":"https://pith.science/paper/KCBKT6XZ","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2108.04217&json=true","fetch_graph":"https://pith.science/api/pith-number/KCBKT6XZ3WVGDOD3OF6UCNHA2Q/graph.json","fetch_events":"https://pith.science/api/pith-number/KCBKT6XZ3WVGDOD3OF6UCNHA2Q/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/KCBKT6XZ3WVGDOD3OF6UCNHA2Q/action/timestamp_anchor","attest_storage":"https://pith.science/pith/KCBKT6XZ3WVGDOD3OF6UCNHA2Q/action/storage_attestation","attest_author":"https://pith.science/pith/KCBKT6XZ3WVGDOD3OF6UCNHA2Q/action/author_attestation","sign_citation":"https://pith.science/pith/KCBKT6XZ3WVGDOD3OF6UCNHA2Q/action/citation_signature","submit_replication":"https://pith.science/pith/KCBKT6XZ3WVGDOD3OF6UCNHA2Q/action/replication_record"}},"created_at":"2026-07-05T03:04:13.888779+00:00","updated_at":"2026-07-05T03:04:13.888779+00:00"}