{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2024:VC4JEIZLEP36IFLTXU4BL6GVKE","short_pith_number":"pith:VC4JEIZL","canonical_record":{"source":{"id":"2412.16475","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","primary_cat":"cs.LG","submitted_at":"2024-12-21T04:07:17Z","cross_cats_sorted":["cs.AI","stat.ML"],"title_canon_sha256":"33cc06efed653c51ab4ccab24d4034696c2009a204f829222342f299743ad2b7","abstract_canon_sha256":"0420c187b6e04a69b4fa4e2b58de702878352ba7d5b29d212a005fe04c13eac3"},"schema_version":"1.0"},"canonical_sha256":"a8b892232b23f7e41573bd3815f8d551260e8156e19991f2444aceefd34a8ec4","source":{"kind":"arxiv","id":"2412.16475","version":1},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2412.16475","created_at":"2026-07-05T09:52:48Z"},{"alias_kind":"arxiv_version","alias_value":"2412.16475v1","created_at":"2026-07-05T09:52:48Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2412.16475","created_at":"2026-07-05T09:52:48Z"},{"alias_kind":"pith_short_12","alias_value":"VC4JEIZLEP36","created_at":"2026-07-05T09:52:48Z"},{"alias_kind":"pith_short_16","alias_value":"VC4JEIZLEP36IFLT","created_at":"2026-07-05T09:52:48Z"},{"alias_kind":"pith_short_8","alias_value":"VC4JEIZL","created_at":"2026-07-05T09:52:48Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2024:VC4JEIZLEP36IFLTXU4BL6GVKE","target":"record","payload":{"canonical_record":{"source":{"id":"2412.16475","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","primary_cat":"cs.LG","submitted_at":"2024-12-21T04:07:17Z","cross_cats_sorted":["cs.AI","stat.ML"],"title_canon_sha256":"33cc06efed653c51ab4ccab24d4034696c2009a204f829222342f299743ad2b7","abstract_canon_sha256":"0420c187b6e04a69b4fa4e2b58de702878352ba7d5b29d212a005fe04c13eac3"},"schema_version":"1.0"},"canonical_sha256":"a8b892232b23f7e41573bd3815f8d551260e8156e19991f2444aceefd34a8ec4","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T09:52:48.394381Z","signature_b64":"g1/DN2qMOllH+TLVzKwMRIirNd6hL+gwpPun65UW1jLLaedj1VS45J6qH9tnyI2w06L3kDjPZT9BzouiONrlBg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"a8b892232b23f7e41573bd3815f8d551260e8156e19991f2444aceefd34a8ec4","last_reissued_at":"2026-07-05T09:52:48.393811Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T09:52:48.393811Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2412.16475","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":"2026-07-05T09:52:48Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"KS5Jo7MD07zGb875pKDbqlrKeNSkIlED0zW7MzS04yDLgvmA0axrTYVyaXCm7TjM1ro11W4k12WRJkV6fgAaAA==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-06T19:06:30.751856Z"},"content_sha256":"d9c0e7b93cd868750983867108de7dc693bb915772e79f9d0f7da429dcb194e0","schema_version":"1.0","event_id":"sha256:d9c0e7b93cd868750983867108de7dc693bb915772e79f9d0f7da429dcb194e0"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2024:VC4JEIZLEP36IFLTXU4BL6GVKE","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"When Can Proxies Improve the Sample Complexity of Preference Learning?","license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","headline":"","cross_cats":["cs.AI","stat.ML"],"primary_cat":"cs.LG","authors_text":"Alexander D'Amour, Daniel Augusto de Souza, Matt J. Kusner, Mengyue Yang, Pasquale Minervini, Yuchen Zhu, Zhengyan Shi","submitted_at":"2024-12-21T04:07:17Z","abstract_excerpt":"We address the problem of reward hacking, where maximising a proxy reward does not necessarily increase the true reward. This is a key concern for Large Language Models (LLMs), as they are often fine-tuned on human preferences that may not accurately reflect a true objective. Existing work uses various tricks such as regularisation, tweaks to the reward model, and reward hacking detectors, to limit the influence that such proxy preferences have on a model. Luckily, in many contexts such as medicine, education, and law, a sparse amount of expert data is often available. In these cases, it is of"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2412.16475","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/2412.16475/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":"2026-07-05T09:52:48Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"tSWvJ06lNUkYGGXKlfixCKrdcViYJxrhdWgzxoRHzMr8UnMFNM33ymYDaxOyoLSwxbCIHI3FCPkwBLSDIdA2Dg==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-06T19:06:30.752344Z"},"content_sha256":"7456349c051502396c67779161ff687211eb8c44369d46b1e1c15c1784f81686","schema_version":"1.0","event_id":"sha256:7456349c051502396c67779161ff687211eb8c44369d46b1e1c15c1784f81686"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/VC4JEIZLEP36IFLTXU4BL6GVKE/bundle.json","state_url":"https://pith.science/pith/VC4JEIZLEP36IFLTXU4BL6GVKE/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/VC4JEIZLEP36IFLTXU4BL6GVKE/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":"2026-08-06T19:06:30Z","links":{"resolver":"https://pith.science/pith/VC4JEIZLEP36IFLTXU4BL6GVKE","bundle":"https://pith.science/pith/VC4JEIZLEP36IFLTXU4BL6GVKE/bundle.json","state":"https://pith.science/pith/VC4JEIZLEP36IFLTXU4BL6GVKE/state.json","well_known_bundle":"https://pith.science/.well-known/pith/VC4JEIZLEP36IFLTXU4BL6GVKE/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2024:VC4JEIZLEP36IFLTXU4BL6GVKE","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":"0420c187b6e04a69b4fa4e2b58de702878352ba7d5b29d212a005fe04c13eac3","cross_cats_sorted":["cs.AI","stat.ML"],"license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","primary_cat":"cs.LG","submitted_at":"2024-12-21T04:07:17Z","title_canon_sha256":"33cc06efed653c51ab4ccab24d4034696c2009a204f829222342f299743ad2b7"},"schema_version":"1.0","source":{"id":"2412.16475","kind":"arxiv","version":1}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2412.16475","created_at":"2026-07-05T09:52:48Z"},{"alias_kind":"arxiv_version","alias_value":"2412.16475v1","created_at":"2026-07-05T09:52:48Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2412.16475","created_at":"2026-07-05T09:52:48Z"},{"alias_kind":"pith_short_12","alias_value":"VC4JEIZLEP36","created_at":"2026-07-05T09:52:48Z"},{"alias_kind":"pith_short_16","alias_value":"VC4JEIZLEP36IFLT","created_at":"2026-07-05T09:52:48Z"},{"alias_kind":"pith_short_8","alias_value":"VC4JEIZL","created_at":"2026-07-05T09:52:48Z"}],"graph_snapshots":[{"event_id":"sha256:7456349c051502396c67779161ff687211eb8c44369d46b1e1c15c1784f81686","target":"graph","created_at":"2026-07-05T09:52:48Z","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/2412.16475/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"We address the problem of reward hacking, where maximising a proxy reward does not necessarily increase the true reward. This is a key concern for Large Language Models (LLMs), as they are often fine-tuned on human preferences that may not accurately reflect a true objective. Existing work uses various tricks such as regularisation, tweaks to the reward model, and reward hacking detectors, to limit the influence that such proxy preferences have on a model. Luckily, in many contexts such as medicine, education, and law, a sparse amount of expert data is often available. In these cases, it is of","authors_text":"Alexander D'Amour, Daniel Augusto de Souza, Matt J. Kusner, Mengyue Yang, Pasquale Minervini, Yuchen Zhu, Zhengyan Shi","cross_cats":["cs.AI","stat.ML"],"headline":"","license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","primary_cat":"cs.LG","submitted_at":"2024-12-21T04:07:17Z","title":"When Can Proxies Improve the Sample Complexity of Preference Learning?"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2412.16475","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:d9c0e7b93cd868750983867108de7dc693bb915772e79f9d0f7da429dcb194e0","target":"record","created_at":"2026-07-05T09:52:48Z","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":"0420c187b6e04a69b4fa4e2b58de702878352ba7d5b29d212a005fe04c13eac3","cross_cats_sorted":["cs.AI","stat.ML"],"license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","primary_cat":"cs.LG","submitted_at":"2024-12-21T04:07:17Z","title_canon_sha256":"33cc06efed653c51ab4ccab24d4034696c2009a204f829222342f299743ad2b7"},"schema_version":"1.0","source":{"id":"2412.16475","kind":"arxiv","version":1}},"canonical_sha256":"a8b892232b23f7e41573bd3815f8d551260e8156e19991f2444aceefd34a8ec4","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"a8b892232b23f7e41573bd3815f8d551260e8156e19991f2444aceefd34a8ec4","first_computed_at":"2026-07-05T09:52:48.393811Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-07-05T09:52:48.393811Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"g1/DN2qMOllH+TLVzKwMRIirNd6hL+gwpPun65UW1jLLaedj1VS45J6qH9tnyI2w06L3kDjPZT9BzouiONrlBg==","signature_status":"signed_v1","signed_at":"2026-07-05T09:52:48.394381Z","signed_message":"canonical_sha256_bytes"},"source_id":"2412.16475","source_kind":"arxiv","source_version":1}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:d9c0e7b93cd868750983867108de7dc693bb915772e79f9d0f7da429dcb194e0","sha256:7456349c051502396c67779161ff687211eb8c44369d46b1e1c15c1784f81686"],"state_sha256":"da0d70a12202f4387ca2da9eccc7e07d9126772a3ea21574eaadb3f7081d0766"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"tJsk3R++D9sPjoDYgwstkJ2ghGLXa6k0F9UXbjAEm4wobKYXv5rqYCR5dwfFoxiddJA/aZohoOYeE9PSSolIBg==","signed_message":"bundle_sha256_bytes","signed_at":"2026-08-06T19:06:30.755497Z","bundle_sha256":"854fbe4c393518ac533781e5ff9ae8c50d8062501aa9517f5fba7baf6a5a6236"}}