{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2024:ZSSK3DQDCJECVIV7SSAKL37ZUJ","short_pith_number":"pith:ZSSK3DQD","canonical_record":{"source":{"id":"2409.17401","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2024-09-25T22:20:11Z","cross_cats_sorted":["stat.ML"],"title_canon_sha256":"8cf247e983ecd46481b57ca31c0098398b04d99af8815c0b81414ea9f432564f","abstract_canon_sha256":"dde5d7d76550560c3fcb9a1e1c4b22fc0cec852419d3e5735a6e500909fbb389"},"schema_version":"1.0"},"canonical_sha256":"cca4ad8e0312482aa2bf9480a5eff9a273f81a8ce7859e91354e59808a25c741","source":{"kind":"arxiv","id":"2409.17401","version":2},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2409.17401","created_at":"2026-07-05T10:22:50Z"},{"alias_kind":"arxiv_version","alias_value":"2409.17401v2","created_at":"2026-07-05T10:22:50Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2409.17401","created_at":"2026-07-05T10:22:50Z"},{"alias_kind":"pith_short_12","alias_value":"ZSSK3DQDCJEC","created_at":"2026-07-05T10:22:50Z"},{"alias_kind":"pith_short_16","alias_value":"ZSSK3DQDCJECVIV7","created_at":"2026-07-05T10:22:50Z"},{"alias_kind":"pith_short_8","alias_value":"ZSSK3DQD","created_at":"2026-07-05T10:22:50Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2024:ZSSK3DQDCJECVIV7SSAKL37ZUJ","target":"record","payload":{"canonical_record":{"source":{"id":"2409.17401","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2024-09-25T22:20:11Z","cross_cats_sorted":["stat.ML"],"title_canon_sha256":"8cf247e983ecd46481b57ca31c0098398b04d99af8815c0b81414ea9f432564f","abstract_canon_sha256":"dde5d7d76550560c3fcb9a1e1c4b22fc0cec852419d3e5735a6e500909fbb389"},"schema_version":"1.0"},"canonical_sha256":"cca4ad8e0312482aa2bf9480a5eff9a273f81a8ce7859e91354e59808a25c741","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:22:50.620926Z","signature_b64":"+kCq/JuIarEsQXEtn7/QI97uEgtzLQioaIrLKfHQIlkopchCiPI8zTChPdIuRdSLQcAtRteBvDlrVnRFO+tEDg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"cca4ad8e0312482aa2bf9480a5eff9a273f81a8ce7859e91354e59808a25c741","last_reissued_at":"2026-07-05T10:22:50.620186Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:22:50.620186Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2409.17401","source_version":2,"attestation_state":"computed"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-07-05T10:22:50Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"HFGGYKzVHMS2JcdQSEizahN2TLObofi6I6mQ6lHcfGSJH8BOhty1KmsOQsbx787uDKKeFUB0zlswj9MfKIOGAQ==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-17T13:56:43.485896Z"},"content_sha256":"712b98115f219b6db26f00168206ab270c426ccf2ed24e0302e2e989ba0c977e","schema_version":"1.0","event_id":"sha256:712b98115f219b6db26f00168206ab270c426ccf2ed24e0302e2e989ba0c977e"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2024:ZSSK3DQDCJECVIV7SSAKL37ZUJ","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"Zeroth-Order Policy Gradient for Reinforcement Learning from Human Feedback without Reward Inference","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["stat.ML"],"primary_cat":"cs.LG","authors_text":"Lei Ying, Qining Zhang","submitted_at":"2024-09-25T22:20:11Z","abstract_excerpt":"Reward inference (learning a reward model from human preferences) is a critical intermediate step in the Reinforcement Learning from Human Feedback (RLHF) pipeline for fine-tuning Large Language Models (LLMs). In practice, RLHF faces fundamental challenges such as distribution shift, reward model overfitting, and problem misspecification. An alternative approach is direct policy optimization without reward inference, such as Direct Preference Optimization (DPO), which provides a much simpler pipeline and has shown empirical success in LLM applications. However, DPO utilizes the closed-form exp"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2409.17401","kind":"arxiv","version":2},"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/2409.17401/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":"2026-07-05T10:22:50Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"Ma3esofsbCOQ7zij0AWka+Om9vkbcf+I/tUQ/a7ta1Cg8eSNOcpCeZKgQRAuWRXDAp/Xo7R2Jj0ySxfVbYK5Dw==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-17T13:56:43.486925Z"},"content_sha256":"f815b52c6a2c32633ad05a01a74133e610922fafb14168e74c077aa2d4150179","schema_version":"1.0","event_id":"sha256:f815b52c6a2c32633ad05a01a74133e610922fafb14168e74c077aa2d4150179"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/ZSSK3DQDCJECVIV7SSAKL37ZUJ/bundle.json","state_url":"https://pith.science/pith/ZSSK3DQDCJECVIV7SSAKL37ZUJ/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/ZSSK3DQDCJECVIV7SSAKL37ZUJ/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":"2026-08-17T13:56:43Z","links":{"resolver":"https://pith.science/pith/ZSSK3DQDCJECVIV7SSAKL37ZUJ","bundle":"https://pith.science/pith/ZSSK3DQDCJECVIV7SSAKL37ZUJ/bundle.json","state":"https://pith.science/pith/ZSSK3DQDCJECVIV7SSAKL37ZUJ/state.json","well_known_bundle":"https://pith.science/.well-known/pith/ZSSK3DQDCJECVIV7SSAKL37ZUJ/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2024:ZSSK3DQDCJECVIV7SSAKL37ZUJ","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":"dde5d7d76550560c3fcb9a1e1c4b22fc0cec852419d3e5735a6e500909fbb389","cross_cats_sorted":["stat.ML"],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2024-09-25T22:20:11Z","title_canon_sha256":"8cf247e983ecd46481b57ca31c0098398b04d99af8815c0b81414ea9f432564f"},"schema_version":"1.0","source":{"id":"2409.17401","kind":"arxiv","version":2}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2409.17401","created_at":"2026-07-05T10:22:50Z"},{"alias_kind":"arxiv_version","alias_value":"2409.17401v2","created_at":"2026-07-05T10:22:50Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2409.17401","created_at":"2026-07-05T10:22:50Z"},{"alias_kind":"pith_short_12","alias_value":"ZSSK3DQDCJEC","created_at":"2026-07-05T10:22:50Z"},{"alias_kind":"pith_short_16","alias_value":"ZSSK3DQDCJECVIV7","created_at":"2026-07-05T10:22:50Z"},{"alias_kind":"pith_short_8","alias_value":"ZSSK3DQD","created_at":"2026-07-05T10:22:50Z"}],"graph_snapshots":[{"event_id":"sha256:f815b52c6a2c32633ad05a01a74133e610922fafb14168e74c077aa2d4150179","target":"graph","created_at":"2026-07-05T10:22:50Z","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/2409.17401/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Reward inference (learning a reward model from human preferences) is a critical intermediate step in the Reinforcement Learning from Human Feedback (RLHF) pipeline for fine-tuning Large Language Models (LLMs). In practice, RLHF faces fundamental challenges such as distribution shift, reward model overfitting, and problem misspecification. An alternative approach is direct policy optimization without reward inference, such as Direct Preference Optimization (DPO), which provides a much simpler pipeline and has shown empirical success in LLM applications. However, DPO utilizes the closed-form exp","authors_text":"Lei Ying, Qining Zhang","cross_cats":["stat.ML"],"headline":"","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2024-09-25T22:20:11Z","title":"Zeroth-Order Policy Gradient for Reinforcement Learning from Human Feedback without Reward Inference"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2409.17401","kind":"arxiv","version":2},"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":"sha256:712b98115f219b6db26f00168206ab270c426ccf2ed24e0302e2e989ba0c977e","target":"record","created_at":"2026-07-05T10:22:50Z","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":"dde5d7d76550560c3fcb9a1e1c4b22fc0cec852419d3e5735a6e500909fbb389","cross_cats_sorted":["stat.ML"],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2024-09-25T22:20:11Z","title_canon_sha256":"8cf247e983ecd46481b57ca31c0098398b04d99af8815c0b81414ea9f432564f"},"schema_version":"1.0","source":{"id":"2409.17401","kind":"arxiv","version":2}},"canonical_sha256":"cca4ad8e0312482aa2bf9480a5eff9a273f81a8ce7859e91354e59808a25c741","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"cca4ad8e0312482aa2bf9480a5eff9a273f81a8ce7859e91354e59808a25c741","first_computed_at":"2026-07-05T10:22:50.620186Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-07-05T10:22:50.620186Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"+kCq/JuIarEsQXEtn7/QI97uEgtzLQioaIrLKfHQIlkopchCiPI8zTChPdIuRdSLQcAtRteBvDlrVnRFO+tEDg==","signature_status":"signed_v1","signed_at":"2026-07-05T10:22:50.620926Z","signed_message":"canonical_sha256_bytes"},"source_id":"2409.17401","source_kind":"arxiv","source_version":2}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:712b98115f219b6db26f00168206ab270c426ccf2ed24e0302e2e989ba0c977e","sha256:f815b52c6a2c32633ad05a01a74133e610922fafb14168e74c077aa2d4150179"],"state_sha256":"1826b57e1e0a6f05da2b9aba1dbab5403e84a1faf846edff5e682f3db6d06044"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"Wk08aVtEbldMycIP8Gh029eSpbtaCWuglT+06zQSpkCjw20BmAUE/XEIivGt1btpgeib/HjiAc74fbzv6g+GBg==","signed_message":"bundle_sha256_bytes","signed_at":"2026-08-17T13:56:43.671151Z","bundle_sha256":"9b9d492130b5b9b5718b5959b72b0f8209228567dd74348f7fccc558a6bf9f8c"}}