{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2023:UF4XFXGQJ47ZTTBJ7V3QN72S6Y","short_pith_number":"pith:UF4XFXGQ","canonical_record":{"source":{"id":"2311.07723","kind":"arxiv","version":3},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.AI","submitted_at":"2023-11-13T20:07:36Z","cross_cats_sorted":["cs.CL","cs.LG"],"title_canon_sha256":"6969a3d1a381716927f63df3679410c8d6c7e94d88e47ab9c9b363bebcc3aa48","abstract_canon_sha256":"7ab211e7c42a0cadad9011f7e61efc29d24add8bdb73768e0d272607b67d6d16"},"schema_version":"1.0"},"canonical_sha256":"a17972dcd04f3f99cc29fd7706ff52f63ad28670f96d3b87da3638dfd055c765","source":{"kind":"arxiv","id":"2311.07723","version":3},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2311.07723","created_at":"2026-07-05T07:24:56Z"},{"alias_kind":"arxiv_version","alias_value":"2311.07723v3","created_at":"2026-07-05T07:24:56Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2311.07723","created_at":"2026-07-05T07:24:56Z"},{"alias_kind":"pith_short_12","alias_value":"UF4XFXGQJ47Z","created_at":"2026-07-05T07:24:56Z"},{"alias_kind":"pith_short_16","alias_value":"UF4XFXGQJ47ZTTBJ","created_at":"2026-07-05T07:24:56Z"},{"alias_kind":"pith_short_8","alias_value":"UF4XFXGQ","created_at":"2026-07-05T07:24:56Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2023:UF4XFXGQJ47ZTTBJ7V3QN72S6Y","target":"record","payload":{"canonical_record":{"source":{"id":"2311.07723","kind":"arxiv","version":3},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.AI","submitted_at":"2023-11-13T20:07:36Z","cross_cats_sorted":["cs.CL","cs.LG"],"title_canon_sha256":"6969a3d1a381716927f63df3679410c8d6c7e94d88e47ab9c9b363bebcc3aa48","abstract_canon_sha256":"7ab211e7c42a0cadad9011f7e61efc29d24add8bdb73768e0d272607b67d6d16"},"schema_version":"1.0"},"canonical_sha256":"a17972dcd04f3f99cc29fd7706ff52f63ad28670f96d3b87da3638dfd055c765","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T07:24:56.468788Z","signature_b64":"rDtNjaSewTunQSq+zZvFIMp/ni8fmGCUdPMlF+2VbcU6tdINm2KS94DSEyvwISeHYFgEm/8/ZNTd2dEbCDnWDg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"a17972dcd04f3f99cc29fd7706ff52f63ad28670f96d3b87da3638dfd055c765","last_reissued_at":"2026-07-05T07:24:56.468295Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T07:24:56.468295Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2311.07723","source_version":3,"attestation_state":"computed"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-07-05T07:24:56Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"8ZpKzyzlRG7GQohobXapwWfBq22IPh+Nlb6SIj/Eq5Qiuei73/qupOFMzcIM41lwCeRNzhM72FlMPuAZX6UyDg==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-17T01:11:00.654486Z"},"content_sha256":"fee6f342d643ba90a167f731d22631caf67e3166185392b4663e4e841c178503","schema_version":"1.0","event_id":"sha256:fee6f342d643ba90a167f731d22631caf67e3166185392b4663e4e841c178503"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2023:UF4XFXGQJ47ZTTBJ7V3QN72S6Y","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"Generalization Analogies: A Testbed for Generalizing AI Oversight to Hard-To-Measure Domains","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CL","cs.LG"],"primary_cat":"cs.AI","authors_text":"Garrett Baker, Joshua Clymer, Rohan Subramani, Sam Wang","submitted_at":"2023-11-13T20:07:36Z","abstract_excerpt":"As AI systems become more intelligent and their behavior becomes more challenging to assess, they may learn to game the flaws of human feedback instead of genuinely striving to follow instructions; however, this risk can be mitigated by controlling how LLMs generalize human feedback to situations where it is unreliable. To better understand how reward models generalize, we craft 69 distribution shifts spanning 8 categories. We find that reward models do not learn to evaluate `instruction-following' by default and instead favor personas that resemble internet text. Techniques for interpreting r"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2311.07723","kind":"arxiv","version":3},"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/2311.07723/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":"2026-07-05T07:24:56Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"Yn9PDPkMYiIfb3X44vH30M3jIlV7zBVT3sAtLCG4Kq3SLBT2H/rjxPZ9F1/SE0sCL+PaX2s3k/sxhsmfSrheBQ==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-17T01:11:00.654993Z"},"content_sha256":"d873fb3bf4924a44480ab3cd84fc4071f78bee855132a0f1d488cf9a6a33f765","schema_version":"1.0","event_id":"sha256:d873fb3bf4924a44480ab3cd84fc4071f78bee855132a0f1d488cf9a6a33f765"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/UF4XFXGQJ47ZTTBJ7V3QN72S6Y/bundle.json","state_url":"https://pith.science/pith/UF4XFXGQJ47ZTTBJ7V3QN72S6Y/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/UF4XFXGQJ47ZTTBJ7V3QN72S6Y/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":"2026-08-17T01:11:00Z","links":{"resolver":"https://pith.science/pith/UF4XFXGQJ47ZTTBJ7V3QN72S6Y","bundle":"https://pith.science/pith/UF4XFXGQJ47ZTTBJ7V3QN72S6Y/bundle.json","state":"https://pith.science/pith/UF4XFXGQJ47ZTTBJ7V3QN72S6Y/state.json","well_known_bundle":"https://pith.science/.well-known/pith/UF4XFXGQJ47ZTTBJ7V3QN72S6Y/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2023:UF4XFXGQJ47ZTTBJ7V3QN72S6Y","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":"7ab211e7c42a0cadad9011f7e61efc29d24add8bdb73768e0d272607b67d6d16","cross_cats_sorted":["cs.CL","cs.LG"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.AI","submitted_at":"2023-11-13T20:07:36Z","title_canon_sha256":"6969a3d1a381716927f63df3679410c8d6c7e94d88e47ab9c9b363bebcc3aa48"},"schema_version":"1.0","source":{"id":"2311.07723","kind":"arxiv","version":3}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2311.07723","created_at":"2026-07-05T07:24:56Z"},{"alias_kind":"arxiv_version","alias_value":"2311.07723v3","created_at":"2026-07-05T07:24:56Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2311.07723","created_at":"2026-07-05T07:24:56Z"},{"alias_kind":"pith_short_12","alias_value":"UF4XFXGQJ47Z","created_at":"2026-07-05T07:24:56Z"},{"alias_kind":"pith_short_16","alias_value":"UF4XFXGQJ47ZTTBJ","created_at":"2026-07-05T07:24:56Z"},{"alias_kind":"pith_short_8","alias_value":"UF4XFXGQ","created_at":"2026-07-05T07:24:56Z"}],"graph_snapshots":[{"event_id":"sha256:d873fb3bf4924a44480ab3cd84fc4071f78bee855132a0f1d488cf9a6a33f765","target":"graph","created_at":"2026-07-05T07:24:56Z","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/2311.07723/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"As AI systems become more intelligent and their behavior becomes more challenging to assess, they may learn to game the flaws of human feedback instead of genuinely striving to follow instructions; however, this risk can be mitigated by controlling how LLMs generalize human feedback to situations where it is unreliable. To better understand how reward models generalize, we craft 69 distribution shifts spanning 8 categories. We find that reward models do not learn to evaluate `instruction-following' by default and instead favor personas that resemble internet text. Techniques for interpreting r","authors_text":"Garrett Baker, Joshua Clymer, Rohan Subramani, Sam Wang","cross_cats":["cs.CL","cs.LG"],"headline":"","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.AI","submitted_at":"2023-11-13T20:07:36Z","title":"Generalization Analogies: A Testbed for Generalizing AI Oversight to Hard-To-Measure Domains"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2311.07723","kind":"arxiv","version":3},"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:fee6f342d643ba90a167f731d22631caf67e3166185392b4663e4e841c178503","target":"record","created_at":"2026-07-05T07:24:56Z","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":"7ab211e7c42a0cadad9011f7e61efc29d24add8bdb73768e0d272607b67d6d16","cross_cats_sorted":["cs.CL","cs.LG"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.AI","submitted_at":"2023-11-13T20:07:36Z","title_canon_sha256":"6969a3d1a381716927f63df3679410c8d6c7e94d88e47ab9c9b363bebcc3aa48"},"schema_version":"1.0","source":{"id":"2311.07723","kind":"arxiv","version":3}},"canonical_sha256":"a17972dcd04f3f99cc29fd7706ff52f63ad28670f96d3b87da3638dfd055c765","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"a17972dcd04f3f99cc29fd7706ff52f63ad28670f96d3b87da3638dfd055c765","first_computed_at":"2026-07-05T07:24:56.468295Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-07-05T07:24:56.468295Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"rDtNjaSewTunQSq+zZvFIMp/ni8fmGCUdPMlF+2VbcU6tdINm2KS94DSEyvwISeHYFgEm/8/ZNTd2dEbCDnWDg==","signature_status":"signed_v1","signed_at":"2026-07-05T07:24:56.468788Z","signed_message":"canonical_sha256_bytes"},"source_id":"2311.07723","source_kind":"arxiv","source_version":3}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:fee6f342d643ba90a167f731d22631caf67e3166185392b4663e4e841c178503","sha256:d873fb3bf4924a44480ab3cd84fc4071f78bee855132a0f1d488cf9a6a33f765"],"state_sha256":"39980b95f24e04b4b361b37ec60035fb96df37c69459aa911b6928d2e546d20b"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"u3nDrCAseZFBSN2uV1EUtzV/HPHeNkBZuU/PFMEYs4j5+bT/LmRwj6d5xMeQ/ewMpFjC+QMsuCZosYaIgl5PAw==","signed_message":"bundle_sha256_bytes","signed_at":"2026-08-17T01:11:00.658174Z","bundle_sha256":"04f178581f6161b258507fefb43031f718d858b9576948b478344010f4a22073"}}