{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:4QSCSFX7AIPS47T4SUFWRUZO77","short_pith_number":"pith:4QSCSFX7","schema_version":"1.0","canonical_sha256":"e4242916ff021f2e7e7c950b68d32efff50f4c1d761768e9c33c32df22f75f61","source":{"kind":"arxiv","id":"2403.01251","version":3},"attestation_state":"computed","paper":{"title":"Accelerating Greedy Coordinate Gradient and General Prompt Optimization via Probe Sampling","license":"http://creativecommons.org/licenses/by-sa/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Anirudh Goyal, Kenji Kawaguchi, Michael Shieh, Tianle Cai, Wenyue Zheng, Xuan Long Do, Yiran Zhao","submitted_at":"2024-03-02T16:23:44Z","abstract_excerpt":"Safety of Large Language Models (LLMs) has become a critical issue given their rapid progresses. Greedy Coordinate Gradient (GCG) is shown to be effective in constructing adversarial prompts to break the aligned LLMs, but optimization of GCG is time-consuming. To reduce the time cost of GCG and enable more comprehensive studies of LLM safety, in this work, we study a new algorithm called $\\texttt{Probe sampling}$. At the core of the algorithm is a mechanism that dynamically determines how similar a smaller draft model's predictions are to the target model's predictions for prompt candidates. W"},"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":"2403.01251","kind":"arxiv","version":3},"metadata":{"license":"http://creativecommons.org/licenses/by-sa/4.0/","primary_cat":"cs.CL","submitted_at":"2024-03-02T16:23:44Z","cross_cats_sorted":[],"title_canon_sha256":"219495cd39bd22e80bcdbc062a757afa04bf7b6fbbe9657a3711699de4c28e6e","abstract_canon_sha256":"22b7d9fe720352fd593f7b94ded9bf3c546f13111da7af86684b35ec2687ee46"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T09:32:42.501422Z","signature_b64":"BcwM/oqqMSF2HV0sCd6Hf7vcAZONleOCQncJfyeaEfKExXQPeUJdwNtxhA2EbjSqKYHQXPQCRstXC6A1Vf8yAA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"e4242916ff021f2e7e7c950b68d32efff50f4c1d761768e9c33c32df22f75f61","last_reissued_at":"2026-07-05T09:32:42.500323Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T09:32:42.500323Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Accelerating Greedy Coordinate Gradient and General Prompt Optimization via Probe Sampling","license":"http://creativecommons.org/licenses/by-sa/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Anirudh Goyal, Kenji Kawaguchi, Michael Shieh, Tianle Cai, Wenyue Zheng, Xuan Long Do, Yiran Zhao","submitted_at":"2024-03-02T16:23:44Z","abstract_excerpt":"Safety of Large Language Models (LLMs) has become a critical issue given their rapid progresses. Greedy Coordinate Gradient (GCG) is shown to be effective in constructing adversarial prompts to break the aligned LLMs, but optimization of GCG is time-consuming. To reduce the time cost of GCG and enable more comprehensive studies of LLM safety, in this work, we study a new algorithm called $\\texttt{Probe sampling}$. At the core of the algorithm is a mechanism that dynamically determines how similar a smaller draft model's predictions are to the target model's predictions for prompt candidates. W"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2403.01251","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/2403.01251/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":"2403.01251","created_at":"2026-07-05T09:32:42.500390+00:00"},{"alias_kind":"arxiv_version","alias_value":"2403.01251v3","created_at":"2026-07-05T09:32:42.500390+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2403.01251","created_at":"2026-07-05T09:32:42.500390+00:00"},{"alias_kind":"pith_short_12","alias_value":"4QSCSFX7AIPS","created_at":"2026-07-05T09:32:42.500390+00:00"},{"alias_kind":"pith_short_16","alias_value":"4QSCSFX7AIPS47T4","created_at":"2026-07-05T09:32:42.500390+00:00"},{"alias_kind":"pith_short_8","alias_value":"4QSCSFX7","created_at":"2026-07-05T09:32:42.500390+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":3,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2410.15362","citing_title":"Faster-GCG: Efficient Discrete Optimization Jailbreak Attacks against Aligned Large Language Models","ref_index":11,"is_internal_anchor":false},{"citing_arxiv_id":"2511.02356","citing_title":"ASTRA: An Automated Framework for Strategy Discovery, Retrieval, and Evolution for Jailbreaking LLMs","ref_index":59,"is_internal_anchor":false},{"citing_arxiv_id":"2605.12813","citing_title":"REALISTA: Realistic Latent Adversarial Attacks that Elicit LLM Hallucinations","ref_index":66,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/4QSCSFX7AIPS47T4SUFWRUZO77","json":"https://pith.science/pith/4QSCSFX7AIPS47T4SUFWRUZO77.json","graph_json":"https://pith.science/api/pith-number/4QSCSFX7AIPS47T4SUFWRUZO77/graph.json","events_json":"https://pith.science/api/pith-number/4QSCSFX7AIPS47T4SUFWRUZO77/events.json","paper":"https://pith.science/paper/4QSCSFX7"},"agent_actions":{"view_html":"https://pith.science/pith/4QSCSFX7AIPS47T4SUFWRUZO77","download_json":"https://pith.science/pith/4QSCSFX7AIPS47T4SUFWRUZO77.json","view_paper":"https://pith.science/paper/4QSCSFX7","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2403.01251&json=true","fetch_graph":"https://pith.science/api/pith-number/4QSCSFX7AIPS47T4SUFWRUZO77/graph.json","fetch_events":"https://pith.science/api/pith-number/4QSCSFX7AIPS47T4SUFWRUZO77/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/4QSCSFX7AIPS47T4SUFWRUZO77/action/timestamp_anchor","attest_storage":"https://pith.science/pith/4QSCSFX7AIPS47T4SUFWRUZO77/action/storage_attestation","attest_author":"https://pith.science/pith/4QSCSFX7AIPS47T4SUFWRUZO77/action/author_attestation","sign_citation":"https://pith.science/pith/4QSCSFX7AIPS47T4SUFWRUZO77/action/citation_signature","submit_replication":"https://pith.science/pith/4QSCSFX7AIPS47T4SUFWRUZO77/action/replication_record"}},"created_at":"2026-07-05T09:32:42.500390+00:00","updated_at":"2026-07-05T09:32:42.500390+00:00"}