{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2023:QZ6YBEA2YFU45H6U5ILPY3HGX2","short_pith_number":"pith:QZ6YBEA2","schema_version":"1.0","canonical_sha256":"867d80901ac169ce9fd4ea16fc6ce6bebf5e0b529918320b12a3a58b649a6f10","source":{"kind":"arxiv","id":"2310.10077","version":1},"attestation_state":"computed","paper":{"title":"Prompt Packer: Deceiving LLMs through Compositional Instruction with Hidden Attacks","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Rui Tang, Shuyu Jiang, Xingshu Chen","submitted_at":"2023-10-16T05:19:25Z","abstract_excerpt":"Recently, Large language models (LLMs) with powerful general capabilities have been increasingly integrated into various Web applications, while undergoing alignment training to ensure that the generated content aligns with user intent and ethics. Unfortunately, they remain the risk of generating harmful content like hate speech and criminal activities in practical applications. Current approaches primarily rely on detecting, collecting, and training against harmful prompts to prevent such risks. However, they typically focused on the \"superficial\" harmful prompts with a solitary intent, ignor"},"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":"2310.10077","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2023-10-16T05:19:25Z","cross_cats_sorted":[],"title_canon_sha256":"211c40c40d7366a37722c16f96dd3a20bc3b3f3a6840fbe31cab62b4285b9c5e","abstract_canon_sha256":"d892eab0d970341f8617d7f857ed4f23b44b2fa24a4be44eee155ad7cf2030db"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T07:01:14.908710Z","signature_b64":"g4JKIEt8BzdJQX7E8hC+GxMNQ/Y2BnAJhRaE/F8ujipFoVMyFGWq9HOnITK7d7ZEpkFe1qERHKWwmbmcHt8CAw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"867d80901ac169ce9fd4ea16fc6ce6bebf5e0b529918320b12a3a58b649a6f10","last_reissued_at":"2026-07-05T07:01:14.908261Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T07:01:14.908261Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Prompt Packer: Deceiving LLMs through Compositional Instruction with Hidden Attacks","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Rui Tang, Shuyu Jiang, Xingshu Chen","submitted_at":"2023-10-16T05:19:25Z","abstract_excerpt":"Recently, Large language models (LLMs) with powerful general capabilities have been increasingly integrated into various Web applications, while undergoing alignment training to ensure that the generated content aligns with user intent and ethics. Unfortunately, they remain the risk of generating harmful content like hate speech and criminal activities in practical applications. Current approaches primarily rely on detecting, collecting, and training against harmful prompts to prevent such risks. However, they typically focused on the \"superficial\" harmful prompts with a solitary intent, ignor"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2310.10077","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/2310.10077/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":"2310.10077","created_at":"2026-07-05T07:01:14.908320+00:00"},{"alias_kind":"arxiv_version","alias_value":"2310.10077v1","created_at":"2026-07-05T07:01:14.908320+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2310.10077","created_at":"2026-07-05T07:01:14.908320+00:00"},{"alias_kind":"pith_short_12","alias_value":"QZ6YBEA2YFU4","created_at":"2026-07-05T07:01:14.908320+00:00"},{"alias_kind":"pith_short_16","alias_value":"QZ6YBEA2YFU45H6U","created_at":"2026-07-05T07:01:14.908320+00:00"},{"alias_kind":"pith_short_8","alias_value":"QZ6YBEA2","created_at":"2026-07-05T07:01:14.908320+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":3,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.24981","citing_title":"Large Language Model Selection with Limited Annotations","ref_index":52,"is_internal_anchor":false},{"citing_arxiv_id":"2605.30650","citing_title":"When AI Meets Wall Street: A Survey on Trustworthy AI in Fintech","ref_index":49,"is_internal_anchor":false},{"citing_arxiv_id":"2412.05579","citing_title":"LLMs-as-Judges: A Comprehensive Survey on LLM-based Evaluation Methods","ref_index":104,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/QZ6YBEA2YFU45H6U5ILPY3HGX2","json":"https://pith.science/pith/QZ6YBEA2YFU45H6U5ILPY3HGX2.json","graph_json":"https://pith.science/api/pith-number/QZ6YBEA2YFU45H6U5ILPY3HGX2/graph.json","events_json":"https://pith.science/api/pith-number/QZ6YBEA2YFU45H6U5ILPY3HGX2/events.json","paper":"https://pith.science/paper/QZ6YBEA2"},"agent_actions":{"view_html":"https://pith.science/pith/QZ6YBEA2YFU45H6U5ILPY3HGX2","download_json":"https://pith.science/pith/QZ6YBEA2YFU45H6U5ILPY3HGX2.json","view_paper":"https://pith.science/paper/QZ6YBEA2","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2310.10077&json=true","fetch_graph":"https://pith.science/api/pith-number/QZ6YBEA2YFU45H6U5ILPY3HGX2/graph.json","fetch_events":"https://pith.science/api/pith-number/QZ6YBEA2YFU45H6U5ILPY3HGX2/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/QZ6YBEA2YFU45H6U5ILPY3HGX2/action/timestamp_anchor","attest_storage":"https://pith.science/pith/QZ6YBEA2YFU45H6U5ILPY3HGX2/action/storage_attestation","attest_author":"https://pith.science/pith/QZ6YBEA2YFU45H6U5ILPY3HGX2/action/author_attestation","sign_citation":"https://pith.science/pith/QZ6YBEA2YFU45H6U5ILPY3HGX2/action/citation_signature","submit_replication":"https://pith.science/pith/QZ6YBEA2YFU45H6U5ILPY3HGX2/action/replication_record"}},"created_at":"2026-07-05T07:01:14.908320+00:00","updated_at":"2026-07-05T07:01:14.908320+00:00"}