{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2026:WA4HPUYAFG4223QK47MXKIF6RE","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":"f99ea8b91d69c7290d1d304f8b51b02d60a37b0c55433ed60a44d39faf7ffc6f","cross_cats_sorted":["cs.AI","cs.CR"],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2026-04-17T09:03:11Z","title_canon_sha256":"7c00df388e9485eac7a07815c9502594e2c8e4d5b1d3171f8a9a695f793a4c53"},"schema_version":"1.0","source":{"id":"2604.15851","kind":"arxiv","version":3}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2604.15851","created_at":"2026-05-20T00:04:31Z"},{"alias_kind":"arxiv_version","alias_value":"2604.15851v3","created_at":"2026-05-20T00:04:31Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2604.15851","created_at":"2026-05-20T00:04:31Z"},{"alias_kind":"pith_short_12","alias_value":"WA4HPUYAFG42","created_at":"2026-05-20T00:04:31Z"},{"alias_kind":"pith_short_16","alias_value":"WA4HPUYAFG4223QK","created_at":"2026-05-20T00:04:31Z"},{"alias_kind":"pith_short_8","alias_value":"WA4HPUYA","created_at":"2026-05-20T00:04:31Z"}],"graph_snapshots":[{"event_id":"sha256:6001f084c557aaa4ed5fb7a08b991adf278b4268dd2dfb54a785dba019984fd2","target":"graph","created_at":"2026-05-20T00:04:31Z","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":3,"items":[{"attestation":"unclaimed","claim_id":"C1","kind":"strongest_claim","source":"verdict.strongest_claim","status":"machine_extracted","text":"Experiments show that while the strongest models handle textbook mechanisms well, all models struggle with advanced algorithms, revealing substantial gaps in current DP reasoning capabilities."},{"attestation":"unclaimed","claim_id":"C2","kind":"weakest_assumption","source":"verdict.weakest_assumption","status":"machine_extracted","text":"The benchmark instances are free of shortcut reasoning patterns and correctly labeled as to whether each function satisfies the stated DP guarantee under the given assumptions."},{"attestation":"unclaimed","claim_id":"C3","kind":"one_line_summary","source":"verdict.one_line_summary","status":"machine_extracted","text":"DPrivBench shows that top LLMs handle basic differential privacy mechanisms but fail on advanced algorithms, exposing gaps in automated DP reasoning."}],"snapshot_sha256":"bc06401d08860c7ce4046d956490a23b1fb65525ac3995b77a1e288b72dcbe24"},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"integrity":{"available":true,"clean":true,"detectors_run":[],"endpoint":"/pith/2604.15851/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Differential privacy (DP) has a wide range of applications for protecting data privacy, but designing and verifying DP algorithms requires expert-level reasoning, creating a high barrier for non-expert practitioners. Prior works either rely on specialized verification languages that demand substantial domain expertise or remain semi-automated and require human-in-the-loop guidance. In this work, we investigate whether large language models (LLMs) can automate DP reasoning. We introduce DPrivBench, a benchmark in which each instance asks whether a function or algorithm satisfies a stated DP gua","authors_text":"Eli Chien, Erchi Wang, Kamalika Chaudhuri, Om Thakkar, Pengrun Huang, Ruihan Wu, Yu-Xiang Wang","cross_cats":["cs.AI","cs.CR"],"headline":"DPrivBench shows that top LLMs handle basic differential privacy mechanisms but fail on advanced algorithms, exposing gaps in automated DP reasoning.","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2026-04-17T09:03:11Z","title":"DPrivBench: Benchmarking LLMs' Reasoning for Differential Privacy"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2604.15851","kind":"arxiv","version":3},"verdict":{"created_at":"2026-05-10T08:50:47.718349Z","id":"12abc099-38f1-4d72-9bc4-c7ee43de3d34","model_set":{"reader":"grok-4.3"},"one_line_summary":"DPrivBench shows that top LLMs handle basic differential privacy mechanisms but fail on advanced algorithms, exposing gaps in automated DP reasoning.","pipeline_version":"pith-pipeline@v0.9.0","pith_extraction_headline":"","strongest_claim":"Experiments show that while the strongest models handle textbook mechanisms well, all models struggle with advanced algorithms, revealing substantial gaps in current DP reasoning capabilities.","weakest_assumption":"The benchmark instances are free of shortcut reasoning patterns and correctly labeled as to whether each function satisfies the stated DP guarantee under the given assumptions."}},"verdict_id":"12abc099-38f1-4d72-9bc4-c7ee43de3d34"}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:c83ce5b8e8b1b12adc11142fac5e8d70f9777b1b57d3c4da8d1c376850bfcb58","target":"record","created_at":"2026-05-20T00:04:31Z","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":"f99ea8b91d69c7290d1d304f8b51b02d60a37b0c55433ed60a44d39faf7ffc6f","cross_cats_sorted":["cs.AI","cs.CR"],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2026-04-17T09:03:11Z","title_canon_sha256":"7c00df388e9485eac7a07815c9502594e2c8e4d5b1d3171f8a9a695f793a4c53"},"schema_version":"1.0","source":{"id":"2604.15851","kind":"arxiv","version":3}},"canonical_sha256":"b03877d30029b9ad6e0ae7d97520be891bddaf91662d0832f3909e11b356cf50","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"b03877d30029b9ad6e0ae7d97520be891bddaf91662d0832f3909e11b356cf50","first_computed_at":"2026-05-20T00:04:31.797946Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-05-20T00:04:31.797946Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"D9ojmYg7meMJGS91ObxhcmBjOZ4EfKgv0gjAVDM/syK+6CK3nCqnNygEwRnJOHZaOcPgWkWX6Jaugq8x3souBw==","signature_status":"signed_v1","signed_at":"2026-05-20T00:04:31.798745Z","signed_message":"canonical_sha256_bytes"},"source_id":"2604.15851","source_kind":"arxiv","source_version":3}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:c83ce5b8e8b1b12adc11142fac5e8d70f9777b1b57d3c4da8d1c376850bfcb58","sha256:6001f084c557aaa4ed5fb7a08b991adf278b4268dd2dfb54a785dba019984fd2"],"state_sha256":"caf43adf944701ae93f442f4628912aea891f0d49359d3d60e0860ecb3f8c3ab"}