{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2026:4CQASH7YCU66ITJ4K32AG2VBH4","short_pith_number":"pith:4CQASH7Y","canonical_record":{"source":{"id":"2608.04322","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2026-08-05T01:05:13Z","cross_cats_sorted":[],"title_canon_sha256":"1d800eeec36125da4f4478cf1cb40100688d24b143d0a5483d10c5bc85c85ae7","abstract_canon_sha256":"939f62779a99b27a9a0d8aa1e85c722c71b3604407339d4edc969e5eead60c28"},"schema_version":"1.0"},"canonical_sha256":"e0a0091ff8153de44d3c56f4036aa13f1d98c9bd553074b77929b91f8bbcf7de","source":{"kind":"arxiv","id":"2608.04322","version":1},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2608.04322","created_at":"2026-08-06T01:32:41Z"},{"alias_kind":"arxiv_version","alias_value":"2608.04322v1","created_at":"2026-08-06T01:32:41Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2608.04322","created_at":"2026-08-06T01:32:41Z"},{"alias_kind":"pith_short_12","alias_value":"4CQASH7YCU66","created_at":"2026-08-06T01:32:41Z"},{"alias_kind":"pith_short_16","alias_value":"4CQASH7YCU66ITJ4","created_at":"2026-08-06T01:32:41Z"},{"alias_kind":"pith_short_8","alias_value":"4CQASH7Y","created_at":"2026-08-06T01:32:41Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2026:4CQASH7YCU66ITJ4K32AG2VBH4","target":"record","payload":{"canonical_record":{"source":{"id":"2608.04322","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2026-08-05T01:05:13Z","cross_cats_sorted":[],"title_canon_sha256":"1d800eeec36125da4f4478cf1cb40100688d24b143d0a5483d10c5bc85c85ae7","abstract_canon_sha256":"939f62779a99b27a9a0d8aa1e85c722c71b3604407339d4edc969e5eead60c28"},"schema_version":"1.0"},"canonical_sha256":"e0a0091ff8153de44d3c56f4036aa13f1d98c9bd553074b77929b91f8bbcf7de","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-08-06T01:32:41.712604Z","signature_b64":"5aRwU7hQRkgkDd+uqgVICIMRuFYSuS/syt/HnbzvrVCCTGMGq2UZxCBVri5qNLRZo42ZR3aPEq0+6nCwbkJMAQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"e0a0091ff8153de44d3c56f4036aa13f1d98c9bd553074b77929b91f8bbcf7de","last_reissued_at":"2026-08-06T01:32:41.711158Z","signature_status":"signed_v1","first_computed_at":"2026-08-06T01:32:41.711158Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2608.04322","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":"2026-08-06T01:32:41Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"iG2pDN71kzh/eTrcFcX0cdw0Z4+n7LuqcaCA1xpDgDT8hstA/PJ1KWvaJTuGkcx8DgeIwWAA4vzJNeq2xfxABw==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-10T17:06:12.277166Z"},"content_sha256":"f6d7cf55d9e0a2f75955ec4f7a8546eb718fe213dfe2f991bd71899204bf6d1f","schema_version":"1.0","event_id":"sha256:f6d7cf55d9e0a2f75955ec4f7a8546eb718fe213dfe2f991bd71899204bf6d1f"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2026:4CQASH7YCU66ITJ4K32AG2VBH4","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"DataRx: Missingness-Aware Sampling for Safer Large Language Model Task-Specific Fine-Tuning","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Junbo Zhang, Qianli Zhou, Wen Jiang, Xinyang Deng","submitted_at":"2026-08-05T01:05:13Z","abstract_excerpt":"Task-specific fine-tuning can improve the performance of large language models (LLMs) on downstream tasks. However, our study reveals that task-specific fine-tuning can also weaken the safety guardrails of aligned LLMs. A widely adopted strategy for preserving safety during fine-tuning is to incorporate safety data. Although previous studies have shown that randomly mixing safety data can alleviate safety degradation, the underlying principle determining why some safety examples are more effective than others still remains unclear. In this paper, we propose DataRx, a missingness-aware sampling"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2608.04322","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/2608.04322/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":"2026-08-06T01:32:41Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"vMOM2EYqw0CnZmhUW9MSK4bld6wYmwW1IxIRjTQgAi/JE5M4uBXpyVWVkMDekn5iEe4BNpB1sgBHX8xmS4vtCg==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-10T17:06:12.278181Z"},"content_sha256":"c86f9fdb0a3ec4aa5ef7badaedaf82041669adfb71810aeddc5b02656044db3f","schema_version":"1.0","event_id":"sha256:c86f9fdb0a3ec4aa5ef7badaedaf82041669adfb71810aeddc5b02656044db3f"},{"event_type":"integrity_finding","subject_pith_number":"pith:2026:4CQASH7YCU66ITJ4K32AG2VBH4","target":"integrity","payload":{"note":"DOI in the printed bibliography is fragmented by whitespace or line breaks. A longer candidate (10.18653/v1/2025.findings-ijcnlp.75) was visible in the surrounding text but could not be confirmed against doi.org as printed.","snippet":"R. V . Tomar, P. Nakov, and Y . Wang, “Unsafechain: Enhancing reasoning model safety via hard cases,” inProceedings of the 14th International Joint Conference on Natural Language Processing and the 4th Conference of the Asia-Pacific Chapter","arxiv_id":"2608.04322","detector":"doi_compliance","evidence":{"ref_index":58,"verdict_class":"incontrovertible","resolved_title":null,"printed_excerpt":"10.18653/v1/2025","reconstructed_doi":"10.18653/v1/2025.findings-ijcnlp.75"},"severity":"advisory","ref_index":58,"audited_at":"2026-08-08T19:58:28.285364Z","event_type":"pith.integrity.v1","detected_doi":"10.18653/v1/2025.findings-ijcnlp.75","detector_url":"https://pith.science/pith-integrity-protocol#doi_compliance","external_url":null,"finding_type":"recoverable_identifier","evidence_hash":"f3b05e1bcbcd5fdb23bd723cc8321e47581d4e0c13f5aac58027ed79e8fcf7d3","paper_version":1,"verdict_class":"incontrovertible","resolved_title":null,"detector_version":"1.1.0","detected_arxiv_id":null,"integrity_event_id":18723,"payload_sha256":"8c0867e08ec299cc5009476ec97ebd79228c47561e3cbefce51ecb1cde164c21","signature_b64":"VDNvsMkEh3nxTwJOprHYXeIbcBntoTqmfvRDz9aW2RVy9kT8gN0OQjVOGTBj/TrBKklTmWFcQ52GXS6dCMc9Ag==","signing_key_id":"pith-v1-2026-05"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-08-08T19:58:33Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"9LXJ00FeHIPqzfnOSY95H+m3Vb5U5LAS0bMmoMClIOwEhiJeKebYyb1+p+4Fm8CP4RHWKa8ju3/QcEq2sc/nDg==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-10T17:06:12.283063Z"},"content_sha256":"9bc25e029722bcc05e7f5961264fd8e9e91b8b5ed29cac6e91c6f534d3352678","schema_version":"1.0","event_id":"sha256:9bc25e029722bcc05e7f5961264fd8e9e91b8b5ed29cac6e91c6f534d3352678"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/4CQASH7YCU66ITJ4K32AG2VBH4/bundle.json","state_url":"https://pith.science/pith/4CQASH7YCU66ITJ4K32AG2VBH4/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/4CQASH7YCU66ITJ4K32AG2VBH4/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":"2026-08-10T17:06:12Z","links":{"resolver":"https://pith.science/pith/4CQASH7YCU66ITJ4K32AG2VBH4","bundle":"https://pith.science/pith/4CQASH7YCU66ITJ4K32AG2VBH4/bundle.json","state":"https://pith.science/pith/4CQASH7YCU66ITJ4K32AG2VBH4/state.json","well_known_bundle":"https://pith.science/.well-known/pith/4CQASH7YCU66ITJ4K32AG2VBH4/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2026:4CQASH7YCU66ITJ4K32AG2VBH4","merge_version":"pith-open-graph-merge-v1","event_count":3,"valid_event_count":3,"invalid_event_count":0,"equivocation_count":0,"current":{"canonical_record":{"metadata":{"abstract_canon_sha256":"939f62779a99b27a9a0d8aa1e85c722c71b3604407339d4edc969e5eead60c28","cross_cats_sorted":[],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2026-08-05T01:05:13Z","title_canon_sha256":"1d800eeec36125da4f4478cf1cb40100688d24b143d0a5483d10c5bc85c85ae7"},"schema_version":"1.0","source":{"id":"2608.04322","kind":"arxiv","version":1}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2608.04322","created_at":"2026-08-06T01:32:41Z"},{"alias_kind":"arxiv_version","alias_value":"2608.04322v1","created_at":"2026-08-06T01:32:41Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2608.04322","created_at":"2026-08-06T01:32:41Z"},{"alias_kind":"pith_short_12","alias_value":"4CQASH7YCU66","created_at":"2026-08-06T01:32:41Z"},{"alias_kind":"pith_short_16","alias_value":"4CQASH7YCU66ITJ4","created_at":"2026-08-06T01:32:41Z"},{"alias_kind":"pith_short_8","alias_value":"4CQASH7Y","created_at":"2026-08-06T01:32:41Z"}],"graph_snapshots":[{"event_id":"sha256:c86f9fdb0a3ec4aa5ef7badaedaf82041669adfb71810aeddc5b02656044db3f","target":"graph","created_at":"2026-08-06T01:32:41Z","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/2608.04322/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Task-specific fine-tuning can improve the performance of large language models (LLMs) on downstream tasks. However, our study reveals that task-specific fine-tuning can also weaken the safety guardrails of aligned LLMs. A widely adopted strategy for preserving safety during fine-tuning is to incorporate safety data. Although previous studies have shown that randomly mixing safety data can alleviate safety degradation, the underlying principle determining why some safety examples are more effective than others still remains unclear. In this paper, we propose DataRx, a missingness-aware sampling","authors_text":"Junbo Zhang, Qianli Zhou, Wen Jiang, Xinyang Deng","cross_cats":[],"headline":"","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2026-08-05T01:05:13Z","title":"DataRx: Missingness-Aware Sampling for Safer Large Language Model Task-Specific Fine-Tuning"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2608.04322","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:f6d7cf55d9e0a2f75955ec4f7a8546eb718fe213dfe2f991bd71899204bf6d1f","target":"record","created_at":"2026-08-06T01:32:41Z","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":"939f62779a99b27a9a0d8aa1e85c722c71b3604407339d4edc969e5eead60c28","cross_cats_sorted":[],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2026-08-05T01:05:13Z","title_canon_sha256":"1d800eeec36125da4f4478cf1cb40100688d24b143d0a5483d10c5bc85c85ae7"},"schema_version":"1.0","source":{"id":"2608.04322","kind":"arxiv","version":1}},"canonical_sha256":"e0a0091ff8153de44d3c56f4036aa13f1d98c9bd553074b77929b91f8bbcf7de","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"e0a0091ff8153de44d3c56f4036aa13f1d98c9bd553074b77929b91f8bbcf7de","first_computed_at":"2026-08-06T01:32:41.711158Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-08-06T01:32:41.711158Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"5aRwU7hQRkgkDd+uqgVICIMRuFYSuS/syt/HnbzvrVCCTGMGq2UZxCBVri5qNLRZo42ZR3aPEq0+6nCwbkJMAQ==","signature_status":"signed_v1","signed_at":"2026-08-06T01:32:41.712604Z","signed_message":"canonical_sha256_bytes"},"source_id":"2608.04322","source_kind":"arxiv","source_version":1}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:f6d7cf55d9e0a2f75955ec4f7a8546eb718fe213dfe2f991bd71899204bf6d1f","sha256:c86f9fdb0a3ec4aa5ef7badaedaf82041669adfb71810aeddc5b02656044db3f","sha256:9bc25e029722bcc05e7f5961264fd8e9e91b8b5ed29cac6e91c6f534d3352678"],"state_sha256":"5fb5c382d6c3b167c2341045756798d9625bf44ba26d6bb3f7bcfb67bb15f73c"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"/xf/VNlZaGA7Tgvv3rlVfRpk/DInlz74SLvdQX7RIyTrASd/RG5vFWEpDE2J2u4HezPRFWE0sUkwp+0TBnKACQ==","signed_message":"bundle_sha256_bytes","signed_at":"2026-08-10T17:06:12.285800Z","bundle_sha256":"7419bc74e1fc92842dc2c66092af0b6a13dc04c25d46c112531c2f108da50ea0"}}