{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2023:Z7766IMZUXBFH4VM7SOBFE64GY","short_pith_number":"pith:Z7766IMZ","canonical_record":{"source":{"id":"2311.17391","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2023-11-29T06:42:36Z","cross_cats_sorted":[],"title_canon_sha256":"f6934f95d2828967d16c5086432a65aaceb62c1b09ba0ae159d4ba3bf84e7e1f","abstract_canon_sha256":"4ab4536bc2f66b57e17066e4793c3ed161ef0ab4bc9cf93635d3893b66c623fd"},"schema_version":"1.0"},"canonical_sha256":"cfffef2199a5c253f2acfc9c1293dc36135ef7719ac4ceb3de47c52c18fa60d0","source":{"kind":"arxiv","id":"2311.17391","version":1},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2311.17391","created_at":"2026-07-05T07:18:02Z"},{"alias_kind":"arxiv_version","alias_value":"2311.17391v1","created_at":"2026-07-05T07:18:02Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2311.17391","created_at":"2026-07-05T07:18:02Z"},{"alias_kind":"pith_short_12","alias_value":"Z7766IMZUXBF","created_at":"2026-07-05T07:18:02Z"},{"alias_kind":"pith_short_16","alias_value":"Z7766IMZUXBFH4VM","created_at":"2026-07-05T07:18:02Z"},{"alias_kind":"pith_short_8","alias_value":"Z7766IMZ","created_at":"2026-07-05T07:18:02Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2023:Z7766IMZUXBFH4VM7SOBFE64GY","target":"record","payload":{"canonical_record":{"source":{"id":"2311.17391","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2023-11-29T06:42:36Z","cross_cats_sorted":[],"title_canon_sha256":"f6934f95d2828967d16c5086432a65aaceb62c1b09ba0ae159d4ba3bf84e7e1f","abstract_canon_sha256":"4ab4536bc2f66b57e17066e4793c3ed161ef0ab4bc9cf93635d3893b66c623fd"},"schema_version":"1.0"},"canonical_sha256":"cfffef2199a5c253f2acfc9c1293dc36135ef7719ac4ceb3de47c52c18fa60d0","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T07:18:02.644540Z","signature_b64":"XgNGpmmG1HwLnkXUPC7G5Rd4ozmSS2Hg3fLjjODblUc52cXbTFn84PIZPG8nrnYiX78Fxq3mcEdhgBjm0EgVDA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"cfffef2199a5c253f2acfc9c1293dc36135ef7719ac4ceb3de47c52c18fa60d0","last_reissued_at":"2026-07-05T07:18:02.644057Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T07:18:02.644057Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2311.17391","source_version":1,"attestation_state":"computed"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-07-05T07:18:02Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"XiQJ9ImXsk/xIKDNFiVqe9bODnBFATGkP6ZM4HGvmgIQc3/lBFyi0lNv/hW2J7ouHZ3+NYi+Bh7rmpmwiXcbAQ==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-07-10T16:47:11.692736Z"},"content_sha256":"e24c262501eaf090dd1d21b8fb0ec1d0586e63757df382824fb8cd9996e0bab6","schema_version":"1.0","event_id":"sha256:e24c262501eaf090dd1d21b8fb0ec1d0586e63757df382824fb8cd9996e0bab6"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2023:Z7766IMZUXBFH4VM7SOBFE64GY","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"Unveiling the Implicit Toxicity in Large Language Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Chengfei Li, Hao Sun, Jiaxin Wen, Jinfeng Bai, Minlie Huang, Pei Ke, Zhexin Zhang","submitted_at":"2023-11-29T06:42:36Z","abstract_excerpt":"The open-endedness of large language models (LLMs) combined with their impressive capabilities may lead to new safety issues when being exploited for malicious use. While recent studies primarily focus on probing toxic outputs that can be easily detected with existing toxicity classifiers, we show that LLMs can generate diverse implicit toxic outputs that are exceptionally difficult to detect via simply zero-shot prompting. Moreover, we propose a reinforcement learning (RL) based attacking method to further induce the implicit toxicity in LLMs. Specifically, we optimize the language model with"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2311.17391","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/2311.17391/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":"2026-07-05T07:18:02Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"pBtvwXutQzy7Jmq55kiLmKR6+BFM/DQTmPXKyDR7D+zSAFRVjk7HorukCU+vH3MocjVV4Qj9abw0XWZbFAVtDg==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-07-10T16:47:11.693150Z"},"content_sha256":"80276d980fa394067b110f87e543872322cc79f3f6059913e697c8873aa51019","schema_version":"1.0","event_id":"sha256:80276d980fa394067b110f87e543872322cc79f3f6059913e697c8873aa51019"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/Z7766IMZUXBFH4VM7SOBFE64GY/bundle.json","state_url":"https://pith.science/pith/Z7766IMZUXBFH4VM7SOBFE64GY/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/Z7766IMZUXBFH4VM7SOBFE64GY/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":"2026-07-10T16:47:11Z","links":{"resolver":"https://pith.science/pith/Z7766IMZUXBFH4VM7SOBFE64GY","bundle":"https://pith.science/pith/Z7766IMZUXBFH4VM7SOBFE64GY/bundle.json","state":"https://pith.science/pith/Z7766IMZUXBFH4VM7SOBFE64GY/state.json","well_known_bundle":"https://pith.science/.well-known/pith/Z7766IMZUXBFH4VM7SOBFE64GY/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2023:Z7766IMZUXBFH4VM7SOBFE64GY","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":"4ab4536bc2f66b57e17066e4793c3ed161ef0ab4bc9cf93635d3893b66c623fd","cross_cats_sorted":[],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2023-11-29T06:42:36Z","title_canon_sha256":"f6934f95d2828967d16c5086432a65aaceb62c1b09ba0ae159d4ba3bf84e7e1f"},"schema_version":"1.0","source":{"id":"2311.17391","kind":"arxiv","version":1}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2311.17391","created_at":"2026-07-05T07:18:02Z"},{"alias_kind":"arxiv_version","alias_value":"2311.17391v1","created_at":"2026-07-05T07:18:02Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2311.17391","created_at":"2026-07-05T07:18:02Z"},{"alias_kind":"pith_short_12","alias_value":"Z7766IMZUXBF","created_at":"2026-07-05T07:18:02Z"},{"alias_kind":"pith_short_16","alias_value":"Z7766IMZUXBFH4VM","created_at":"2026-07-05T07:18:02Z"},{"alias_kind":"pith_short_8","alias_value":"Z7766IMZ","created_at":"2026-07-05T07:18:02Z"}],"graph_snapshots":[{"event_id":"sha256:80276d980fa394067b110f87e543872322cc79f3f6059913e697c8873aa51019","target":"graph","created_at":"2026-07-05T07:18:02Z","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/2311.17391/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"The open-endedness of large language models (LLMs) combined with their impressive capabilities may lead to new safety issues when being exploited for malicious use. While recent studies primarily focus on probing toxic outputs that can be easily detected with existing toxicity classifiers, we show that LLMs can generate diverse implicit toxic outputs that are exceptionally difficult to detect via simply zero-shot prompting. Moreover, we propose a reinforcement learning (RL) based attacking method to further induce the implicit toxicity in LLMs. Specifically, we optimize the language model with","authors_text":"Chengfei Li, Hao Sun, Jiaxin Wen, Jinfeng Bai, Minlie Huang, Pei Ke, Zhexin Zhang","cross_cats":[],"headline":"","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2023-11-29T06:42:36Z","title":"Unveiling the Implicit Toxicity in Large Language Models"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2311.17391","kind":"arxiv","version":1},"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:e24c262501eaf090dd1d21b8fb0ec1d0586e63757df382824fb8cd9996e0bab6","target":"record","created_at":"2026-07-05T07:18:02Z","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":"4ab4536bc2f66b57e17066e4793c3ed161ef0ab4bc9cf93635d3893b66c623fd","cross_cats_sorted":[],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2023-11-29T06:42:36Z","title_canon_sha256":"f6934f95d2828967d16c5086432a65aaceb62c1b09ba0ae159d4ba3bf84e7e1f"},"schema_version":"1.0","source":{"id":"2311.17391","kind":"arxiv","version":1}},"canonical_sha256":"cfffef2199a5c253f2acfc9c1293dc36135ef7719ac4ceb3de47c52c18fa60d0","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"cfffef2199a5c253f2acfc9c1293dc36135ef7719ac4ceb3de47c52c18fa60d0","first_computed_at":"2026-07-05T07:18:02.644057Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-07-05T07:18:02.644057Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"XgNGpmmG1HwLnkXUPC7G5Rd4ozmSS2Hg3fLjjODblUc52cXbTFn84PIZPG8nrnYiX78Fxq3mcEdhgBjm0EgVDA==","signature_status":"signed_v1","signed_at":"2026-07-05T07:18:02.644540Z","signed_message":"canonical_sha256_bytes"},"source_id":"2311.17391","source_kind":"arxiv","source_version":1}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:e24c262501eaf090dd1d21b8fb0ec1d0586e63757df382824fb8cd9996e0bab6","sha256:80276d980fa394067b110f87e543872322cc79f3f6059913e697c8873aa51019"],"state_sha256":"ee4af2fbec44edd059ae50756692d9ac9ba320e99077425e5f635b0dd3b059d2"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"cfcpx09yayuk3yjqZ7eaar/0kKE8l65eCeuEtDEx3Yyb9Xd5fXgUGw2cemxOnq5PJRcKDKAziJTE7c6k0AQ/Bg==","signed_message":"bundle_sha256_bytes","signed_at":"2026-07-10T16:47:11.695135Z","bundle_sha256":"aaec43dffe44f814e7b953144ee6dac556c027483848306796ef2d779dda3598"}}