{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:CVFSLXIWN6T4DUFAKIP6VGGPMD","short_pith_number":"pith:CVFSLXIW","schema_version":"1.0","canonical_sha256":"154b25dd166fa7c1d0a0521fea98cf60cddffe83256d430390c4fc1e2b07706b","source":{"kind":"arxiv","id":"2509.00391","version":1},"attestation_state":"computed","paper":{"title":"The Resurgence of GCG Adversarial Attacks on Large Language Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.CR","cs.LG"],"primary_cat":"cs.CL","authors_text":"Huizhen Shu, Peikang Hu, Xuying Li, Yuting Tan, Zhuo Li","submitted_at":"2025-08-30T07:04:29Z","abstract_excerpt":"Gradient-based adversarial prompting, such as the Greedy Coordinate Gradient (GCG) algorithm, has emerged as a powerful method for jailbreaking large language models (LLMs). In this paper, we present a systematic appraisal of GCG and its annealing-augmented variant, T-GCG, across open-source LLMs of varying scales. Using Qwen2.5-0.5B, LLaMA-3.2-1B, and GPT-OSS-20B, we evaluate attack effectiveness on both safety-oriented prompts (AdvBench) and reasoning-intensive coding prompts. Our study reveals three key findings: (1) attack success rates (ASR) decrease with model size, reflecting the increa"},"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":"2509.00391","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2025-08-30T07:04:29Z","cross_cats_sorted":["cs.AI","cs.CR","cs.LG"],"title_canon_sha256":"18cb1bb359cffd72f2f8221243fd34f4211ee804dc9b35e54004368852ec1c87","abstract_canon_sha256":"3b7828974bb846826035d97293cb1cd8696afa9bf1924bb05d450ebe11379d1f"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T12:02:15.341693Z","signature_b64":"SzfxHMKSQ9Dx8ZERIOtZ/BBd0mdYmrBxK8/ZvjEZN2f369in61OnZzAhjaCUIx/DuHzrSX3BFgrH4K+tEz4QDw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"154b25dd166fa7c1d0a0521fea98cf60cddffe83256d430390c4fc1e2b07706b","last_reissued_at":"2026-07-05T12:02:15.341192Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T12:02:15.341192Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"The Resurgence of GCG Adversarial Attacks on Large Language Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.CR","cs.LG"],"primary_cat":"cs.CL","authors_text":"Huizhen Shu, Peikang Hu, Xuying Li, Yuting Tan, Zhuo Li","submitted_at":"2025-08-30T07:04:29Z","abstract_excerpt":"Gradient-based adversarial prompting, such as the Greedy Coordinate Gradient (GCG) algorithm, has emerged as a powerful method for jailbreaking large language models (LLMs). In this paper, we present a systematic appraisal of GCG and its annealing-augmented variant, T-GCG, across open-source LLMs of varying scales. Using Qwen2.5-0.5B, LLaMA-3.2-1B, and GPT-OSS-20B, we evaluate attack effectiveness on both safety-oriented prompts (AdvBench) and reasoning-intensive coding prompts. Our study reveals three key findings: (1) attack success rates (ASR) decrease with model size, reflecting the increa"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2509.00391","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/2509.00391/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":"2509.00391","created_at":"2026-07-05T12:02:15.341262+00:00"},{"alias_kind":"arxiv_version","alias_value":"2509.00391v1","created_at":"2026-07-05T12:02:15.341262+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2509.00391","created_at":"2026-07-05T12:02:15.341262+00:00"},{"alias_kind":"pith_short_12","alias_value":"CVFSLXIWN6T4","created_at":"2026-07-05T12:02:15.341262+00:00"},{"alias_kind":"pith_short_16","alias_value":"CVFSLXIWN6T4DUFA","created_at":"2026-07-05T12:02:15.341262+00:00"},{"alias_kind":"pith_short_8","alias_value":"CVFSLXIW","created_at":"2026-07-05T12:02:15.341262+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.02946","citing_title":"RouteHijack: Routing-Aware Attack on Mixture-of-Experts LLMs","ref_index":57,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/CVFSLXIWN6T4DUFAKIP6VGGPMD","json":"https://pith.science/pith/CVFSLXIWN6T4DUFAKIP6VGGPMD.json","graph_json":"https://pith.science/api/pith-number/CVFSLXIWN6T4DUFAKIP6VGGPMD/graph.json","events_json":"https://pith.science/api/pith-number/CVFSLXIWN6T4DUFAKIP6VGGPMD/events.json","paper":"https://pith.science/paper/CVFSLXIW"},"agent_actions":{"view_html":"https://pith.science/pith/CVFSLXIWN6T4DUFAKIP6VGGPMD","download_json":"https://pith.science/pith/CVFSLXIWN6T4DUFAKIP6VGGPMD.json","view_paper":"https://pith.science/paper/CVFSLXIW","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2509.00391&json=true","fetch_graph":"https://pith.science/api/pith-number/CVFSLXIWN6T4DUFAKIP6VGGPMD/graph.json","fetch_events":"https://pith.science/api/pith-number/CVFSLXIWN6T4DUFAKIP6VGGPMD/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/CVFSLXIWN6T4DUFAKIP6VGGPMD/action/timestamp_anchor","attest_storage":"https://pith.science/pith/CVFSLXIWN6T4DUFAKIP6VGGPMD/action/storage_attestation","attest_author":"https://pith.science/pith/CVFSLXIWN6T4DUFAKIP6VGGPMD/action/author_attestation","sign_citation":"https://pith.science/pith/CVFSLXIWN6T4DUFAKIP6VGGPMD/action/citation_signature","submit_replication":"https://pith.science/pith/CVFSLXIWN6T4DUFAKIP6VGGPMD/action/replication_record"}},"created_at":"2026-07-05T12:02:15.341262+00:00","updated_at":"2026-07-05T12:02:15.341262+00:00"}