{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:U7W5YITWA6LGFSP3A2OC5J6CDY","short_pith_number":"pith:U7W5YITW","schema_version":"1.0","canonical_sha256":"a7eddc2276079662c9fb069c2ea7c21e32e9841b2f78576c61d00e66fde85e6f","source":{"kind":"arxiv","id":"2404.17120","version":2},"attestation_state":"computed","paper":{"title":"Talking Nonsense: Probing Large Language Models' Understanding of Adversarial Gibberish Inputs","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CL","authors_text":"James Zou, Valeriia Cherepanova","submitted_at":"2024-04-26T02:29:26Z","abstract_excerpt":"Large language models (LLMs) exhibit excellent ability to understand human languages, but do they also understand their own language that appears gibberish to us? In this work we delve into this question, aiming to uncover the mechanisms underlying such behavior in LLMs. We employ the Greedy Coordinate Gradient optimizer to craft prompts that compel LLMs to generate coherent responses from seemingly nonsensical inputs. We call these inputs LM Babel and this work systematically studies the behavior of LLMs manipulated by these prompts. We find that the manipulation efficiency depends on the tar"},"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":"2404.17120","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2024-04-26T02:29:26Z","cross_cats_sorted":["cs.AI","cs.LG"],"title_canon_sha256":"a94c2f80957b8e38ba69c24c6a55203ead4a865fcea84db57c4e9c9462e30f97","abstract_canon_sha256":"9588d728d9a047bd4e33e127fae323bb0ae1120d55a9b744a439049b5c0bbeb5"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T08:13:14.262985Z","signature_b64":"fVtdwj/VqoUcQDKUoI/YX8VBHLz+tXqGrgYFgOGdnHCtiHcnDcM7Pe55JrQBpMC4sMpUgR7+AT8FSnUY2a0IAg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"a7eddc2276079662c9fb069c2ea7c21e32e9841b2f78576c61d00e66fde85e6f","last_reissued_at":"2026-07-05T08:13:14.262547Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T08:13:14.262547Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Talking Nonsense: Probing Large Language Models' Understanding of Adversarial Gibberish Inputs","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CL","authors_text":"James Zou, Valeriia Cherepanova","submitted_at":"2024-04-26T02:29:26Z","abstract_excerpt":"Large language models (LLMs) exhibit excellent ability to understand human languages, but do they also understand their own language that appears gibberish to us? In this work we delve into this question, aiming to uncover the mechanisms underlying such behavior in LLMs. We employ the Greedy Coordinate Gradient optimizer to craft prompts that compel LLMs to generate coherent responses from seemingly nonsensical inputs. We call these inputs LM Babel and this work systematically studies the behavior of LLMs manipulated by these prompts. We find that the manipulation efficiency depends on the tar"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2404.17120","kind":"arxiv","version":2},"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/2404.17120/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":"2404.17120","created_at":"2026-07-05T08:13:14.262608+00:00"},{"alias_kind":"arxiv_version","alias_value":"2404.17120v2","created_at":"2026-07-05T08:13:14.262608+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2404.17120","created_at":"2026-07-05T08:13:14.262608+00:00"},{"alias_kind":"pith_short_12","alias_value":"U7W5YITWA6LG","created_at":"2026-07-05T08:13:14.262608+00:00"},{"alias_kind":"pith_short_16","alias_value":"U7W5YITWA6LGFSP3","created_at":"2026-07-05T08:13:14.262608+00:00"},{"alias_kind":"pith_short_8","alias_value":"U7W5YITW","created_at":"2026-07-05T08:13:14.262608+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2502.09974","citing_title":"Has My System Prompt Been Used? Large Language Model Prompt Membership Inference","ref_index":6,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/U7W5YITWA6LGFSP3A2OC5J6CDY","json":"https://pith.science/pith/U7W5YITWA6LGFSP3A2OC5J6CDY.json","graph_json":"https://pith.science/api/pith-number/U7W5YITWA6LGFSP3A2OC5J6CDY/graph.json","events_json":"https://pith.science/api/pith-number/U7W5YITWA6LGFSP3A2OC5J6CDY/events.json","paper":"https://pith.science/paper/U7W5YITW"},"agent_actions":{"view_html":"https://pith.science/pith/U7W5YITWA6LGFSP3A2OC5J6CDY","download_json":"https://pith.science/pith/U7W5YITWA6LGFSP3A2OC5J6CDY.json","view_paper":"https://pith.science/paper/U7W5YITW","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2404.17120&json=true","fetch_graph":"https://pith.science/api/pith-number/U7W5YITWA6LGFSP3A2OC5J6CDY/graph.json","fetch_events":"https://pith.science/api/pith-number/U7W5YITWA6LGFSP3A2OC5J6CDY/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/U7W5YITWA6LGFSP3A2OC5J6CDY/action/timestamp_anchor","attest_storage":"https://pith.science/pith/U7W5YITWA6LGFSP3A2OC5J6CDY/action/storage_attestation","attest_author":"https://pith.science/pith/U7W5YITWA6LGFSP3A2OC5J6CDY/action/author_attestation","sign_citation":"https://pith.science/pith/U7W5YITWA6LGFSP3A2OC5J6CDY/action/citation_signature","submit_replication":"https://pith.science/pith/U7W5YITWA6LGFSP3A2OC5J6CDY/action/replication_record"}},"created_at":"2026-07-05T08:13:14.262608+00:00","updated_at":"2026-07-05T08:13:14.262608+00:00"}