{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2025:TIMPT5RJUBTXSEDSR5U35S6LBN","short_pith_number":"pith:TIMPT5RJ","canonical_record":{"source":{"id":"2504.16272","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-04-22T21:09:33Z","cross_cats_sorted":[],"title_canon_sha256":"05ceb4c4acc92d2e1d482f6643e2502d18d375475ab3f0c032d1113069907bb9","abstract_canon_sha256":"834fe1b11960f7630db9490ee98ebb78d5641d0b6ca9a3c6db1633f79e3119cd"},"schema_version":"1.0"},"canonical_sha256":"9a18f9f629a0677910728f69becbcb0b5354b4110502d21c303d4824660fd1ff","source":{"kind":"arxiv","id":"2504.16272","version":1},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2504.16272","created_at":"2026-07-05T10:52:44Z"},{"alias_kind":"arxiv_version","alias_value":"2504.16272v1","created_at":"2026-07-05T10:52:44Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2504.16272","created_at":"2026-07-05T10:52:44Z"},{"alias_kind":"pith_short_12","alias_value":"TIMPT5RJUBTX","created_at":"2026-07-05T10:52:44Z"},{"alias_kind":"pith_short_16","alias_value":"TIMPT5RJUBTXSEDS","created_at":"2026-07-05T10:52:44Z"},{"alias_kind":"pith_short_8","alias_value":"TIMPT5RJ","created_at":"2026-07-05T10:52:44Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2025:TIMPT5RJUBTXSEDSR5U35S6LBN","target":"record","payload":{"canonical_record":{"source":{"id":"2504.16272","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-04-22T21:09:33Z","cross_cats_sorted":[],"title_canon_sha256":"05ceb4c4acc92d2e1d482f6643e2502d18d375475ab3f0c032d1113069907bb9","abstract_canon_sha256":"834fe1b11960f7630db9490ee98ebb78d5641d0b6ca9a3c6db1633f79e3119cd"},"schema_version":"1.0"},"canonical_sha256":"9a18f9f629a0677910728f69becbcb0b5354b4110502d21c303d4824660fd1ff","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:52:44.041431Z","signature_b64":"cVT1BiDbyCXpa0aodv8nnl6EjC3nSyLHd3tS/07ddlLm6rqthkI1VOkrkRp8C7Faxf3+EhZ7goMbJEMa/X7lAA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"9a18f9f629a0677910728f69becbcb0b5354b4110502d21c303d4824660fd1ff","last_reissued_at":"2026-07-05T10:52:44.040971Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:52:44.040971Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2504.16272","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":"2026-07-05T10:52:44Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"2Rd3LAzMa+yqNGKB+CLn4MnQO3nl4GSv98KvFNdj0GZLZEgpGQgwtIR5R4gMeUmD6WWAyNd2l0hJndqRPI9QAg==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-16T19:20:45.826762Z"},"content_sha256":"e6563c98b097b1f5be079ef7cfd0ddca002cf28725350bde2a233ec22a042dee","schema_version":"1.0","event_id":"sha256:e6563c98b097b1f5be079ef7cfd0ddca002cf28725350bde2a233ec22a042dee"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2025:TIMPT5RJUBTXSEDSR5U35S6LBN","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"Learning Explainable Dense Reward Shapes via Bayesian Optimization","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Dongyeop Kang, Ian Yang, Kwang-Sung Jun, Mingyi Hong, Ryan Koo, Vipul Raheja","submitted_at":"2025-04-22T21:09:33Z","abstract_excerpt":"Current reinforcement learning from human feedback (RLHF) pipelines for large language model (LLM) alignment typically assign scalar rewards to sequences, using the final token as a surrogate indicator for the quality of the entire sequence. However, this leads to sparse feedback and suboptimal token-level credit assignment. In this work, we frame reward shaping as an optimization problem focused on token-level credit assignment. We propose a reward-shaping function leveraging explainability methods such as SHAP and LIME to estimate per-token rewards from the reward model. To learn parameters "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2504.16272","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/2504.16272/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":"2026-07-05T10:52:44Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"xWB6kiYL8xqzlaPnVuaYBLeYyAR7rtY8Bd/t8ARiTecVbGCdn1lgJWjfq9VnpbG54bIJV8E8l/jUZRTuQoTzCw==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-16T19:20:45.827393Z"},"content_sha256":"60e254d27a47dd4b8318e9f3e56958a508a07cf7e7bc48943729945e0d95f578","schema_version":"1.0","event_id":"sha256:60e254d27a47dd4b8318e9f3e56958a508a07cf7e7bc48943729945e0d95f578"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/TIMPT5RJUBTXSEDSR5U35S6LBN/bundle.json","state_url":"https://pith.science/pith/TIMPT5RJUBTXSEDSR5U35S6LBN/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/TIMPT5RJUBTXSEDSR5U35S6LBN/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":"2026-08-16T19:20:45Z","links":{"resolver":"https://pith.science/pith/TIMPT5RJUBTXSEDSR5U35S6LBN","bundle":"https://pith.science/pith/TIMPT5RJUBTXSEDSR5U35S6LBN/bundle.json","state":"https://pith.science/pith/TIMPT5RJUBTXSEDSR5U35S6LBN/state.json","well_known_bundle":"https://pith.science/.well-known/pith/TIMPT5RJUBTXSEDSR5U35S6LBN/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2025:TIMPT5RJUBTXSEDSR5U35S6LBN","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":"834fe1b11960f7630db9490ee98ebb78d5641d0b6ca9a3c6db1633f79e3119cd","cross_cats_sorted":[],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-04-22T21:09:33Z","title_canon_sha256":"05ceb4c4acc92d2e1d482f6643e2502d18d375475ab3f0c032d1113069907bb9"},"schema_version":"1.0","source":{"id":"2504.16272","kind":"arxiv","version":1}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2504.16272","created_at":"2026-07-05T10:52:44Z"},{"alias_kind":"arxiv_version","alias_value":"2504.16272v1","created_at":"2026-07-05T10:52:44Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2504.16272","created_at":"2026-07-05T10:52:44Z"},{"alias_kind":"pith_short_12","alias_value":"TIMPT5RJUBTX","created_at":"2026-07-05T10:52:44Z"},{"alias_kind":"pith_short_16","alias_value":"TIMPT5RJUBTXSEDS","created_at":"2026-07-05T10:52:44Z"},{"alias_kind":"pith_short_8","alias_value":"TIMPT5RJ","created_at":"2026-07-05T10:52:44Z"}],"graph_snapshots":[{"event_id":"sha256:60e254d27a47dd4b8318e9f3e56958a508a07cf7e7bc48943729945e0d95f578","target":"graph","created_at":"2026-07-05T10:52:44Z","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/2504.16272/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Current reinforcement learning from human feedback (RLHF) pipelines for large language model (LLM) alignment typically assign scalar rewards to sequences, using the final token as a surrogate indicator for the quality of the entire sequence. However, this leads to sparse feedback and suboptimal token-level credit assignment. In this work, we frame reward shaping as an optimization problem focused on token-level credit assignment. We propose a reward-shaping function leveraging explainability methods such as SHAP and LIME to estimate per-token rewards from the reward model. To learn parameters ","authors_text":"Dongyeop Kang, Ian Yang, Kwang-Sung Jun, Mingyi Hong, Ryan Koo, Vipul Raheja","cross_cats":[],"headline":"","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-04-22T21:09:33Z","title":"Learning Explainable Dense Reward Shapes via Bayesian Optimization"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2504.16272","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:e6563c98b097b1f5be079ef7cfd0ddca002cf28725350bde2a233ec22a042dee","target":"record","created_at":"2026-07-05T10:52:44Z","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":"834fe1b11960f7630db9490ee98ebb78d5641d0b6ca9a3c6db1633f79e3119cd","cross_cats_sorted":[],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-04-22T21:09:33Z","title_canon_sha256":"05ceb4c4acc92d2e1d482f6643e2502d18d375475ab3f0c032d1113069907bb9"},"schema_version":"1.0","source":{"id":"2504.16272","kind":"arxiv","version":1}},"canonical_sha256":"9a18f9f629a0677910728f69becbcb0b5354b4110502d21c303d4824660fd1ff","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"9a18f9f629a0677910728f69becbcb0b5354b4110502d21c303d4824660fd1ff","first_computed_at":"2026-07-05T10:52:44.040971Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-07-05T10:52:44.040971Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"cVT1BiDbyCXpa0aodv8nnl6EjC3nSyLHd3tS/07ddlLm6rqthkI1VOkrkRp8C7Faxf3+EhZ7goMbJEMa/X7lAA==","signature_status":"signed_v1","signed_at":"2026-07-05T10:52:44.041431Z","signed_message":"canonical_sha256_bytes"},"source_id":"2504.16272","source_kind":"arxiv","source_version":1}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:e6563c98b097b1f5be079ef7cfd0ddca002cf28725350bde2a233ec22a042dee","sha256:60e254d27a47dd4b8318e9f3e56958a508a07cf7e7bc48943729945e0d95f578"],"state_sha256":"871827429c1f2408adf3b1bc8e831cb715524b3971f936384cf0c43c8f3ff776"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"R6Lm5cC+M8NOp+pY61QUf8EbhUh8pM9UzWTkOq/zQkrWJH50DeCgPVx+sAO/e37Tx9w4WTPa1VHm3k789wY1AQ==","signed_message":"bundle_sha256_bytes","signed_at":"2026-08-16T19:20:45.833299Z","bundle_sha256":"b3d8706daf998c7c8e8ded67c3d27dc2e2813f6c61b4b8e205d925a0609c2499"}}