{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2025:VC6X3BEXBQLESEEOBZBKLAUL73","short_pith_number":"pith:VC6X3BEX","canonical_record":{"source":{"id":"2503.22230","kind":"arxiv","version":3},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-03-28T08:26:41Z","cross_cats_sorted":[],"title_canon_sha256":"9e57e31b4e817387e81a3855790a06a276dacc95bda4c9e0ee5f3c59775f0a50","abstract_canon_sha256":"be2fae74c0e3d4c1f3cb68f816e4cd582a46eef307a6f2bf32564318a119fcfa"},"schema_version":"1.0"},"canonical_sha256":"a8bd7d84970c1649108e0e42a5828bfeca83eab7d67d4ae82d1e7f5ca775347d","source":{"kind":"arxiv","id":"2503.22230","version":3},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2503.22230","created_at":"2026-07-05T10:43:27Z"},{"alias_kind":"arxiv_version","alias_value":"2503.22230v3","created_at":"2026-07-05T10:43:27Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2503.22230","created_at":"2026-07-05T10:43:27Z"},{"alias_kind":"pith_short_12","alias_value":"VC6X3BEXBQLE","created_at":"2026-07-05T10:43:27Z"},{"alias_kind":"pith_short_16","alias_value":"VC6X3BEXBQLESEEO","created_at":"2026-07-05T10:43:27Z"},{"alias_kind":"pith_short_8","alias_value":"VC6X3BEX","created_at":"2026-07-05T10:43:27Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2025:VC6X3BEXBQLESEEOBZBKLAUL73","target":"record","payload":{"canonical_record":{"source":{"id":"2503.22230","kind":"arxiv","version":3},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-03-28T08:26:41Z","cross_cats_sorted":[],"title_canon_sha256":"9e57e31b4e817387e81a3855790a06a276dacc95bda4c9e0ee5f3c59775f0a50","abstract_canon_sha256":"be2fae74c0e3d4c1f3cb68f816e4cd582a46eef307a6f2bf32564318a119fcfa"},"schema_version":"1.0"},"canonical_sha256":"a8bd7d84970c1649108e0e42a5828bfeca83eab7d67d4ae82d1e7f5ca775347d","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:43:27.026272Z","signature_b64":"DacWijrxlxsxxF3FL39AuqG/YQKhEuwhdtZ04kJ2lMlzqX6/hnvqr2iLQUFbjZQRz8rkevNZloTiJpYnAAFtBg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"a8bd7d84970c1649108e0e42a5828bfeca83eab7d67d4ae82d1e7f5ca775347d","last_reissued_at":"2026-07-05T10:43:27.025786Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:43:27.025786Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2503.22230","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":"2026-07-05T10:43:27Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"0lTe2q4LRf/MHkFHMZN4zzZFaDBUYsVYVeiGvBuKmAPzG7VZIvfv4ta04ZXntEMLZ9osdTFpt1oN+W6IiP6dBg==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-07T12:34:30.298601Z"},"content_sha256":"823b4871a62effd9f28115eb914ce830867869be5e23b481c16bfc0fab14193d","schema_version":"1.0","event_id":"sha256:823b4871a62effd9f28115eb914ce830867869be5e23b481c16bfc0fab14193d"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2025:VC6X3BEXBQLESEEOBZBKLAUL73","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"Exploring Data Scaling Trends and Effects in Reinforcement Learning from Human Feedback","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Chao Xin, Guanlin Liu, Lin Yan, Qingping Yang, Ruofei Zhu, Wei Shen, Yu Yue, Zheng Wu","submitted_at":"2025-03-28T08:26:41Z","abstract_excerpt":"Reinforcement Learning from Human Feedback (RLHF) is crucial for aligning large language models with human preferences. While recent research has focused on algorithmic improvements, the importance of prompt-data construction has been overlooked. This paper addresses this gap by exploring data-driven bottlenecks in RLHF performance scaling, particularly reward hacking and decreasing response diversity. We introduce a hybrid reward system combining reasoning task verifiers (RTV) and a generative reward model (GenRM) to mitigate reward hacking. We also propose a novel prompt-selection method, Pr"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2503.22230","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/2503.22230/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":"2026-07-05T10:43:27Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"1XGXiVH1ogwfJE12eEqtNyuU6c2KeP8cR35Hbfe0ikXvsXEKkxP5m1Rxa3G7CwjUrNYvoCfMm5tCdCMDe+jKBg==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-07T12:34:30.299261Z"},"content_sha256":"b4a322e5ca8e4b9148daeaa0fd73175c0ad29c6c8a2c3ccb09353d5a6546a122","schema_version":"1.0","event_id":"sha256:b4a322e5ca8e4b9148daeaa0fd73175c0ad29c6c8a2c3ccb09353d5a6546a122"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/VC6X3BEXBQLESEEOBZBKLAUL73/bundle.json","state_url":"https://pith.science/pith/VC6X3BEXBQLESEEOBZBKLAUL73/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/VC6X3BEXBQLESEEOBZBKLAUL73/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":"2026-08-07T12:34:30Z","links":{"resolver":"https://pith.science/pith/VC6X3BEXBQLESEEOBZBKLAUL73","bundle":"https://pith.science/pith/VC6X3BEXBQLESEEOBZBKLAUL73/bundle.json","state":"https://pith.science/pith/VC6X3BEXBQLESEEOBZBKLAUL73/state.json","well_known_bundle":"https://pith.science/.well-known/pith/VC6X3BEXBQLESEEOBZBKLAUL73/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2025:VC6X3BEXBQLESEEOBZBKLAUL73","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":"be2fae74c0e3d4c1f3cb68f816e4cd582a46eef307a6f2bf32564318a119fcfa","cross_cats_sorted":[],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-03-28T08:26:41Z","title_canon_sha256":"9e57e31b4e817387e81a3855790a06a276dacc95bda4c9e0ee5f3c59775f0a50"},"schema_version":"1.0","source":{"id":"2503.22230","kind":"arxiv","version":3}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2503.22230","created_at":"2026-07-05T10:43:27Z"},{"alias_kind":"arxiv_version","alias_value":"2503.22230v3","created_at":"2026-07-05T10:43:27Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2503.22230","created_at":"2026-07-05T10:43:27Z"},{"alias_kind":"pith_short_12","alias_value":"VC6X3BEXBQLE","created_at":"2026-07-05T10:43:27Z"},{"alias_kind":"pith_short_16","alias_value":"VC6X3BEXBQLESEEO","created_at":"2026-07-05T10:43:27Z"},{"alias_kind":"pith_short_8","alias_value":"VC6X3BEX","created_at":"2026-07-05T10:43:27Z"}],"graph_snapshots":[{"event_id":"sha256:b4a322e5ca8e4b9148daeaa0fd73175c0ad29c6c8a2c3ccb09353d5a6546a122","target":"graph","created_at":"2026-07-05T10:43:27Z","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/2503.22230/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Reinforcement Learning from Human Feedback (RLHF) is crucial for aligning large language models with human preferences. While recent research has focused on algorithmic improvements, the importance of prompt-data construction has been overlooked. This paper addresses this gap by exploring data-driven bottlenecks in RLHF performance scaling, particularly reward hacking and decreasing response diversity. We introduce a hybrid reward system combining reasoning task verifiers (RTV) and a generative reward model (GenRM) to mitigate reward hacking. We also propose a novel prompt-selection method, Pr","authors_text":"Chao Xin, Guanlin Liu, Lin Yan, Qingping Yang, Ruofei Zhu, Wei Shen, Yu Yue, Zheng Wu","cross_cats":[],"headline":"","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-03-28T08:26:41Z","title":"Exploring Data Scaling Trends and Effects in Reinforcement Learning from Human Feedback"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2503.22230","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":"sha256:823b4871a62effd9f28115eb914ce830867869be5e23b481c16bfc0fab14193d","target":"record","created_at":"2026-07-05T10:43:27Z","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":"be2fae74c0e3d4c1f3cb68f816e4cd582a46eef307a6f2bf32564318a119fcfa","cross_cats_sorted":[],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-03-28T08:26:41Z","title_canon_sha256":"9e57e31b4e817387e81a3855790a06a276dacc95bda4c9e0ee5f3c59775f0a50"},"schema_version":"1.0","source":{"id":"2503.22230","kind":"arxiv","version":3}},"canonical_sha256":"a8bd7d84970c1649108e0e42a5828bfeca83eab7d67d4ae82d1e7f5ca775347d","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"a8bd7d84970c1649108e0e42a5828bfeca83eab7d67d4ae82d1e7f5ca775347d","first_computed_at":"2026-07-05T10:43:27.025786Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-07-05T10:43:27.025786Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"DacWijrxlxsxxF3FL39AuqG/YQKhEuwhdtZ04kJ2lMlzqX6/hnvqr2iLQUFbjZQRz8rkevNZloTiJpYnAAFtBg==","signature_status":"signed_v1","signed_at":"2026-07-05T10:43:27.026272Z","signed_message":"canonical_sha256_bytes"},"source_id":"2503.22230","source_kind":"arxiv","source_version":3}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:823b4871a62effd9f28115eb914ce830867869be5e23b481c16bfc0fab14193d","sha256:b4a322e5ca8e4b9148daeaa0fd73175c0ad29c6c8a2c3ccb09353d5a6546a122"],"state_sha256":"b78637a10b13af0a38667878a0a6e6e77b5e81c8ad65b3298f42c3677f09e6e1"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"STEa25CjR/CS1rLQLrjshtGV9w+mcnRDe5HdceOy5PiwovtIU8ctKjH/+Uwgv/cP3q0gPutcYyEKF5gfEjBZAA==","signed_message":"bundle_sha256_bytes","signed_at":"2026-08-07T12:34:30.306784Z","bundle_sha256":"bd272ed2d5fb403292f1041fd3d586ba6ee42465a6b3a173a42390bbdce050fc"}}