{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2026:5MYO7XENOTME45X6FECGIGDZDX","short_pith_number":"pith:5MYO7XEN","canonical_record":{"source":{"id":"2607.23394","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.AI","submitted_at":"2026-07-25T23:41:47Z","cross_cats_sorted":[],"title_canon_sha256":"eaf814fc0d5574e1540ccb40ab20f5569e1b3138ed1c7a0ee91c7670b1bdcce1","abstract_canon_sha256":"9270cc2ef37c82d5ec82c49ed0acab5bfd7b3f604a405cc68c84544f7032d8d6"},"schema_version":"1.0"},"canonical_sha256":"eb30efdc8d74d84e76fe29046418791de18923a950f3655bfbcdb45bb435c03e","source":{"kind":"arxiv","id":"2607.23394","version":1},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2607.23394","created_at":"2026-07-28T01:22:50Z"},{"alias_kind":"arxiv_version","alias_value":"2607.23394v1","created_at":"2026-07-28T01:22:50Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2607.23394","created_at":"2026-07-28T01:22:50Z"},{"alias_kind":"pith_short_12","alias_value":"5MYO7XENOTME","created_at":"2026-07-28T01:22:50Z"},{"alias_kind":"pith_short_16","alias_value":"5MYO7XENOTME45X6","created_at":"2026-07-28T01:22:50Z"},{"alias_kind":"pith_short_8","alias_value":"5MYO7XEN","created_at":"2026-07-28T01:22:50Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2026:5MYO7XENOTME45X6FECGIGDZDX","target":"record","payload":{"canonical_record":{"source":{"id":"2607.23394","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.AI","submitted_at":"2026-07-25T23:41:47Z","cross_cats_sorted":[],"title_canon_sha256":"eaf814fc0d5574e1540ccb40ab20f5569e1b3138ed1c7a0ee91c7670b1bdcce1","abstract_canon_sha256":"9270cc2ef37c82d5ec82c49ed0acab5bfd7b3f604a405cc68c84544f7032d8d6"},"schema_version":"1.0"},"canonical_sha256":"eb30efdc8d74d84e76fe29046418791de18923a950f3655bfbcdb45bb435c03e","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-28T01:22:50.433638Z","signature_b64":"AH9azPiBKTSAksl6mPeiKgXdlCTq5KLdF1AAvFZW3w/nOlTjmEcDFZ3cMBX1lsrcHakXkrtQ9q1mjqusbNK2DA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"eb30efdc8d74d84e76fe29046418791de18923a950f3655bfbcdb45bb435c03e","last_reissued_at":"2026-07-28T01:22:50.432880Z","signature_status":"signed_v1","first_computed_at":"2026-07-28T01:22:50.432880Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2607.23394","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":"2026-07-28T01:22:50Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"/wtvfrQOfjkfVvyZnjHZ3pmzariL8gboiRtQM4lgfd1/RD3vMC1jN1FAwzhG+O+SKwBy/wY5U+7X1Cu+TQFTCw==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-07-31T18:47:33.430785Z"},"content_sha256":"bfbf51f7ef3026481b62deea1d6e35842349769f035f3a483bbb9018ca31216b","schema_version":"1.0","event_id":"sha256:bfbf51f7ef3026481b62deea1d6e35842349769f035f3a483bbb9018ca31216b"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2026:5MYO7XENOTME45X6FECGIGDZDX","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"Inference-Time Consensus for Mitigating Hidden Behaviors from LLM Fine-Tuning","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.AI","authors_text":"Adhyyan Narang, Artin Tajdini, Claire Zhang, Jamie Morgenstern","submitted_at":"2026-07-25T23:41:47Z","abstract_excerpt":"Recent work shows that fine-tuning language models on even a small amount of poisoned data can install targeted misbehavior, and ostensibly benign data can transmit hidden preferences that generalize broadly. Standard defenses, such as data filtering, mixing in harmless data, and regularization, attenuate these effects but do not eliminate them. We instead pursue robustness through redundancy: collecting multiple datasets from different sources and only learning what is common between them. Thus, if only a subset of sources are malicious, the misbehavior will be blocked. In order to implement "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2607.23394","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/2607.23394/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":"2026-07-28T01:22:50Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"JfafDWtI3NZhiZsFzsd0DEQc+62k07jf47llO3MAM/DhpHwN7NBN7Of9FcM9kPxy6eohnnZ/+Nh1v0Ch5MaBBg==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-07-31T18:47:33.431414Z"},"content_sha256":"e3ecc4db26c242f1d10d604dc802789bf069dcaaab126a5a1875863796ac15c8","schema_version":"1.0","event_id":"sha256:e3ecc4db26c242f1d10d604dc802789bf069dcaaab126a5a1875863796ac15c8"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/5MYO7XENOTME45X6FECGIGDZDX/bundle.json","state_url":"https://pith.science/pith/5MYO7XENOTME45X6FECGIGDZDX/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/5MYO7XENOTME45X6FECGIGDZDX/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":"2026-07-31T18:47:33Z","links":{"resolver":"https://pith.science/pith/5MYO7XENOTME45X6FECGIGDZDX","bundle":"https://pith.science/pith/5MYO7XENOTME45X6FECGIGDZDX/bundle.json","state":"https://pith.science/pith/5MYO7XENOTME45X6FECGIGDZDX/state.json","well_known_bundle":"https://pith.science/.well-known/pith/5MYO7XENOTME45X6FECGIGDZDX/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2026:5MYO7XENOTME45X6FECGIGDZDX","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":"9270cc2ef37c82d5ec82c49ed0acab5bfd7b3f604a405cc68c84544f7032d8d6","cross_cats_sorted":[],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.AI","submitted_at":"2026-07-25T23:41:47Z","title_canon_sha256":"eaf814fc0d5574e1540ccb40ab20f5569e1b3138ed1c7a0ee91c7670b1bdcce1"},"schema_version":"1.0","source":{"id":"2607.23394","kind":"arxiv","version":1}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2607.23394","created_at":"2026-07-28T01:22:50Z"},{"alias_kind":"arxiv_version","alias_value":"2607.23394v1","created_at":"2026-07-28T01:22:50Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2607.23394","created_at":"2026-07-28T01:22:50Z"},{"alias_kind":"pith_short_12","alias_value":"5MYO7XENOTME","created_at":"2026-07-28T01:22:50Z"},{"alias_kind":"pith_short_16","alias_value":"5MYO7XENOTME45X6","created_at":"2026-07-28T01:22:50Z"},{"alias_kind":"pith_short_8","alias_value":"5MYO7XEN","created_at":"2026-07-28T01:22:50Z"}],"graph_snapshots":[{"event_id":"sha256:e3ecc4db26c242f1d10d604dc802789bf069dcaaab126a5a1875863796ac15c8","target":"graph","created_at":"2026-07-28T01:22:50Z","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/2607.23394/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Recent work shows that fine-tuning language models on even a small amount of poisoned data can install targeted misbehavior, and ostensibly benign data can transmit hidden preferences that generalize broadly. Standard defenses, such as data filtering, mixing in harmless data, and regularization, attenuate these effects but do not eliminate them. We instead pursue robustness through redundancy: collecting multiple datasets from different sources and only learning what is common between them. Thus, if only a subset of sources are malicious, the misbehavior will be blocked. In order to implement ","authors_text":"Adhyyan Narang, Artin Tajdini, Claire Zhang, Jamie Morgenstern","cross_cats":[],"headline":"","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.AI","submitted_at":"2026-07-25T23:41:47Z","title":"Inference-Time Consensus for Mitigating Hidden Behaviors from LLM Fine-Tuning"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2607.23394","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:bfbf51f7ef3026481b62deea1d6e35842349769f035f3a483bbb9018ca31216b","target":"record","created_at":"2026-07-28T01:22:50Z","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":"9270cc2ef37c82d5ec82c49ed0acab5bfd7b3f604a405cc68c84544f7032d8d6","cross_cats_sorted":[],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.AI","submitted_at":"2026-07-25T23:41:47Z","title_canon_sha256":"eaf814fc0d5574e1540ccb40ab20f5569e1b3138ed1c7a0ee91c7670b1bdcce1"},"schema_version":"1.0","source":{"id":"2607.23394","kind":"arxiv","version":1}},"canonical_sha256":"eb30efdc8d74d84e76fe29046418791de18923a950f3655bfbcdb45bb435c03e","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"eb30efdc8d74d84e76fe29046418791de18923a950f3655bfbcdb45bb435c03e","first_computed_at":"2026-07-28T01:22:50.432880Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-07-28T01:22:50.432880Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"AH9azPiBKTSAksl6mPeiKgXdlCTq5KLdF1AAvFZW3w/nOlTjmEcDFZ3cMBX1lsrcHakXkrtQ9q1mjqusbNK2DA==","signature_status":"signed_v1","signed_at":"2026-07-28T01:22:50.433638Z","signed_message":"canonical_sha256_bytes"},"source_id":"2607.23394","source_kind":"arxiv","source_version":1}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:bfbf51f7ef3026481b62deea1d6e35842349769f035f3a483bbb9018ca31216b","sha256:e3ecc4db26c242f1d10d604dc802789bf069dcaaab126a5a1875863796ac15c8"],"state_sha256":"0446b9311f44a853ddd6a6912b6be02fbb93aee803e9bcecdb49528833204a63"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"d8GwcshV0eKw25aCq5sEAtqJ7u5ejNyKLOiBjVjwHovrrHk8tgSpysOopbKMbb2O4OxTA7Iuy7y3sMp83yI8Cw==","signed_message":"bundle_sha256_bytes","signed_at":"2026-07-31T18:47:33.438933Z","bundle_sha256":"313e92b9f9620de889d95bfae717061e7d86ca27485595e471877c70f1b03a85"}}