{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2022:JH5TUUQQN2BMCXVSY2ESPFL2MQ","short_pith_number":"pith:JH5TUUQQ","schema_version":"1.0","canonical_sha256":"49fb3a52106e82c15eb2c68927957a643ccd3abae5c0a8fb94765eef695fb816","source":{"kind":"arxiv","id":"2205.13621","version":2},"attestation_state":"computed","paper":{"title":"Differentially Private Decoding in Large Language Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CL","authors_text":"Charith Peris, Christophe Dupuy, Jimit Majmudar, Rahul Gupta, Richard Zemel, Sami Smaili","submitted_at":"2022-05-26T20:50:58Z","abstract_excerpt":"Recent large-scale natural language processing (NLP) systems use a pre-trained Large Language Model (LLM) on massive and diverse corpora as a headstart. In practice, the pre-trained model is adapted to a wide array of tasks via fine-tuning on task-specific datasets. LLMs, while effective, have been shown to memorize instances of training data thereby potentially revealing private information processed during pre-training. The potential leakage might further propagate to the downstream tasks for which LLMs are fine-tuned. On the other hand, privacy-preserving algorithms usually involve retraini"},"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":"2205.13621","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2022-05-26T20:50:58Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"c5b85304997f05fca4887540e542655fcbd092c12876c102cf2e58d5bb150098","abstract_canon_sha256":"8fa5603857cadf8a82bf29be22aba7d96b192e6f678a0e525bc468df543a9136"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T04:55:52.656214Z","signature_b64":"4c6hnJYW3tZgu5TDWOiF5gKL+SEVBb7oVDM+YJjpA20dMiX8gusbxqvt0Ngsu7L4dJFrDg4FPvL56TRa2b5yCA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"49fb3a52106e82c15eb2c68927957a643ccd3abae5c0a8fb94765eef695fb816","last_reissued_at":"2026-07-05T04:55:52.655767Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T04:55:52.655767Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Differentially Private Decoding in Large Language Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CL","authors_text":"Charith Peris, Christophe Dupuy, Jimit Majmudar, Rahul Gupta, Richard Zemel, Sami Smaili","submitted_at":"2022-05-26T20:50:58Z","abstract_excerpt":"Recent large-scale natural language processing (NLP) systems use a pre-trained Large Language Model (LLM) on massive and diverse corpora as a headstart. In practice, the pre-trained model is adapted to a wide array of tasks via fine-tuning on task-specific datasets. LLMs, while effective, have been shown to memorize instances of training data thereby potentially revealing private information processed during pre-training. The potential leakage might further propagate to the downstream tasks for which LLMs are fine-tuned. On the other hand, privacy-preserving algorithms usually involve retraini"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2205.13621","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/2205.13621/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":"2205.13621","created_at":"2026-07-05T04:55:52.655831+00:00"},{"alias_kind":"arxiv_version","alias_value":"2205.13621v2","created_at":"2026-07-05T04:55:52.655831+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2205.13621","created_at":"2026-07-05T04:55:52.655831+00:00"},{"alias_kind":"pith_short_12","alias_value":"JH5TUUQQN2BM","created_at":"2026-07-05T04:55:52.655831+00:00"},{"alias_kind":"pith_short_16","alias_value":"JH5TUUQQN2BMCXVS","created_at":"2026-07-05T04:55:52.655831+00:00"},{"alias_kind":"pith_short_8","alias_value":"JH5TUUQQ","created_at":"2026-07-05T04:55:52.655831+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":5,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.03399","citing_title":"Selective Token-Level Cryptographic Redaction for Privacy-Preserving Clinical Deployment of Large Language Models","ref_index":18,"is_internal_anchor":false},{"citing_arxiv_id":"2606.31016","citing_title":"Beyond Wireless Security: Covert Communications in Large Language Model-enabled Edge Networks","ref_index":13,"is_internal_anchor":false},{"citing_arxiv_id":"2511.14045","citing_title":"Auditing Data Membership in Reinforcement Learning With Verifiable Rewards","ref_index":40,"is_internal_anchor":false},{"citing_arxiv_id":"2605.01425","citing_title":"Barriers to Counterfactual Credit Attribution for Autoregressive Models","ref_index":5,"is_internal_anchor":false},{"citing_arxiv_id":"2604.04461","citing_title":"DP-OPD: Differentially Private On-Policy Distillation for Language Models","ref_index":10,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/JH5TUUQQN2BMCXVSY2ESPFL2MQ","json":"https://pith.science/pith/JH5TUUQQN2BMCXVSY2ESPFL2MQ.json","graph_json":"https://pith.science/api/pith-number/JH5TUUQQN2BMCXVSY2ESPFL2MQ/graph.json","events_json":"https://pith.science/api/pith-number/JH5TUUQQN2BMCXVSY2ESPFL2MQ/events.json","paper":"https://pith.science/paper/JH5TUUQQ"},"agent_actions":{"view_html":"https://pith.science/pith/JH5TUUQQN2BMCXVSY2ESPFL2MQ","download_json":"https://pith.science/pith/JH5TUUQQN2BMCXVSY2ESPFL2MQ.json","view_paper":"https://pith.science/paper/JH5TUUQQ","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2205.13621&json=true","fetch_graph":"https://pith.science/api/pith-number/JH5TUUQQN2BMCXVSY2ESPFL2MQ/graph.json","fetch_events":"https://pith.science/api/pith-number/JH5TUUQQN2BMCXVSY2ESPFL2MQ/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/JH5TUUQQN2BMCXVSY2ESPFL2MQ/action/timestamp_anchor","attest_storage":"https://pith.science/pith/JH5TUUQQN2BMCXVSY2ESPFL2MQ/action/storage_attestation","attest_author":"https://pith.science/pith/JH5TUUQQN2BMCXVSY2ESPFL2MQ/action/author_attestation","sign_citation":"https://pith.science/pith/JH5TUUQQN2BMCXVSY2ESPFL2MQ/action/citation_signature","submit_replication":"https://pith.science/pith/JH5TUUQQN2BMCXVSY2ESPFL2MQ/action/replication_record"}},"created_at":"2026-07-05T04:55:52.655831+00:00","updated_at":"2026-07-05T04:55:52.655831+00:00"}