{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2024:AHBRFNOIEN5O2KO4S4FS5ZIUIT","short_pith_number":"pith:AHBRFNOI","canonical_record":{"source":{"id":"2404.00978","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2024-04-01T07:49:11Z","cross_cats_sorted":[],"title_canon_sha256":"30538f4fde870f4ecba4522ff93bdbbf3db1f0289741edcdab0351c191ebfcab","abstract_canon_sha256":"7e00b2bf9d279a3f781fb616f4e7758ef1e363efa4b66f0d763833bedc316dfc"},"schema_version":"1.0"},"canonical_sha256":"01c312b5c8237aed29dc970b2ee51444c363a81e5996f13b0a17512ed0dde28f","source":{"kind":"arxiv","id":"2404.00978","version":2},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2404.00978","created_at":"2026-07-05T09:08:24Z"},{"alias_kind":"arxiv_version","alias_value":"2404.00978v2","created_at":"2026-07-05T09:08:24Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2404.00978","created_at":"2026-07-05T09:08:24Z"},{"alias_kind":"pith_short_12","alias_value":"AHBRFNOIEN5O","created_at":"2026-07-05T09:08:24Z"},{"alias_kind":"pith_short_16","alias_value":"AHBRFNOIEN5O2KO4","created_at":"2026-07-05T09:08:24Z"},{"alias_kind":"pith_short_8","alias_value":"AHBRFNOI","created_at":"2026-07-05T09:08:24Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2024:AHBRFNOIEN5O2KO4S4FS5ZIUIT","target":"record","payload":{"canonical_record":{"source":{"id":"2404.00978","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2024-04-01T07:49:11Z","cross_cats_sorted":[],"title_canon_sha256":"30538f4fde870f4ecba4522ff93bdbbf3db1f0289741edcdab0351c191ebfcab","abstract_canon_sha256":"7e00b2bf9d279a3f781fb616f4e7758ef1e363efa4b66f0d763833bedc316dfc"},"schema_version":"1.0"},"canonical_sha256":"01c312b5c8237aed29dc970b2ee51444c363a81e5996f13b0a17512ed0dde28f","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T09:08:24.365457Z","signature_b64":"q9i9RA6DrNPUWlWPEpxBSmYrLxo6OMAs4FMah1Fki115Kd/wQU13x6PsjdnbePlE2ZQ7Ur48gG1eCmak3XyMDA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"01c312b5c8237aed29dc970b2ee51444c363a81e5996f13b0a17512ed0dde28f","last_reissued_at":"2026-07-05T09:08:24.364931Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T09:08:24.364931Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2404.00978","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":"2026-07-05T09:08:24Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"sgj5sAUXN1ftammJzFZFooIx+mM0w6uFPI4JhSV3Qk0jdpD6q+ec1mu5RRey8tSWyBEfqBODFZZN4kGZ7JNIBg==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-12T05:15:11.051710Z"},"content_sha256":"88c9c3af2f48602e356681e80d0d8b5cd690dcd4d61b40fe034b568a06bde1a8","schema_version":"1.0","event_id":"sha256:88c9c3af2f48602e356681e80d0d8b5cd690dcd4d61b40fe034b568a06bde1a8"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2024:AHBRFNOIEN5O2KO4S4FS5ZIUIT","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"Prior Constraints-based Reward Model Training for Aligning Large Language Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Chenglong Wang, Chunliang Zhang, Hang Zhou, Jingbo Zhu, Tong Xiao, Yimin Hu","submitted_at":"2024-04-01T07:49:11Z","abstract_excerpt":"Reinforcement learning with human feedback for aligning large language models (LLMs) trains a reward model typically using ranking loss with comparison pairs.However, the training procedure suffers from an inherent problem: the uncontrolled scaling of reward scores during reinforcement learning due to the lack of constraints while training the reward model.This paper proposes a Prior Constraints-based Reward Model (namely PCRM) training method to mitigate this problem. PCRM incorporates prior constraints, specifically, length ratio and cosine similarity between outputs of each comparison pair,"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2404.00978","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/2404.00978/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":"2026-07-05T09:08:24Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"5U6ff0uDXEwiLQFQzl84pBnQaVFh2yTzSeQfLQc1R4OJ6XxImj35+HQk2bRS0zyNcPSjYB1swWKknwADh9X0Ag==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-12T05:15:11.052227Z"},"content_sha256":"1330bb053044076f900293313f1374e5d4acdd2045fea224f9c47f763226faa0","schema_version":"1.0","event_id":"sha256:1330bb053044076f900293313f1374e5d4acdd2045fea224f9c47f763226faa0"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/AHBRFNOIEN5O2KO4S4FS5ZIUIT/bundle.json","state_url":"https://pith.science/pith/AHBRFNOIEN5O2KO4S4FS5ZIUIT/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/AHBRFNOIEN5O2KO4S4FS5ZIUIT/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":"2026-08-12T05:15:11Z","links":{"resolver":"https://pith.science/pith/AHBRFNOIEN5O2KO4S4FS5ZIUIT","bundle":"https://pith.science/pith/AHBRFNOIEN5O2KO4S4FS5ZIUIT/bundle.json","state":"https://pith.science/pith/AHBRFNOIEN5O2KO4S4FS5ZIUIT/state.json","well_known_bundle":"https://pith.science/.well-known/pith/AHBRFNOIEN5O2KO4S4FS5ZIUIT/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2024:AHBRFNOIEN5O2KO4S4FS5ZIUIT","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":"7e00b2bf9d279a3f781fb616f4e7758ef1e363efa4b66f0d763833bedc316dfc","cross_cats_sorted":[],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2024-04-01T07:49:11Z","title_canon_sha256":"30538f4fde870f4ecba4522ff93bdbbf3db1f0289741edcdab0351c191ebfcab"},"schema_version":"1.0","source":{"id":"2404.00978","kind":"arxiv","version":2}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2404.00978","created_at":"2026-07-05T09:08:24Z"},{"alias_kind":"arxiv_version","alias_value":"2404.00978v2","created_at":"2026-07-05T09:08:24Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2404.00978","created_at":"2026-07-05T09:08:24Z"},{"alias_kind":"pith_short_12","alias_value":"AHBRFNOIEN5O","created_at":"2026-07-05T09:08:24Z"},{"alias_kind":"pith_short_16","alias_value":"AHBRFNOIEN5O2KO4","created_at":"2026-07-05T09:08:24Z"},{"alias_kind":"pith_short_8","alias_value":"AHBRFNOI","created_at":"2026-07-05T09:08:24Z"}],"graph_snapshots":[{"event_id":"sha256:1330bb053044076f900293313f1374e5d4acdd2045fea224f9c47f763226faa0","target":"graph","created_at":"2026-07-05T09:08:24Z","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/2404.00978/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Reinforcement learning with human feedback for aligning large language models (LLMs) trains a reward model typically using ranking loss with comparison pairs.However, the training procedure suffers from an inherent problem: the uncontrolled scaling of reward scores during reinforcement learning due to the lack of constraints while training the reward model.This paper proposes a Prior Constraints-based Reward Model (namely PCRM) training method to mitigate this problem. PCRM incorporates prior constraints, specifically, length ratio and cosine similarity between outputs of each comparison pair,","authors_text":"Chenglong Wang, Chunliang Zhang, Hang Zhou, Jingbo Zhu, Tong Xiao, Yimin Hu","cross_cats":[],"headline":"","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2024-04-01T07:49:11Z","title":"Prior Constraints-based Reward Model Training for Aligning Large Language Models"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2404.00978","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":"sha256:88c9c3af2f48602e356681e80d0d8b5cd690dcd4d61b40fe034b568a06bde1a8","target":"record","created_at":"2026-07-05T09:08:24Z","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":"7e00b2bf9d279a3f781fb616f4e7758ef1e363efa4b66f0d763833bedc316dfc","cross_cats_sorted":[],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2024-04-01T07:49:11Z","title_canon_sha256":"30538f4fde870f4ecba4522ff93bdbbf3db1f0289741edcdab0351c191ebfcab"},"schema_version":"1.0","source":{"id":"2404.00978","kind":"arxiv","version":2}},"canonical_sha256":"01c312b5c8237aed29dc970b2ee51444c363a81e5996f13b0a17512ed0dde28f","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"01c312b5c8237aed29dc970b2ee51444c363a81e5996f13b0a17512ed0dde28f","first_computed_at":"2026-07-05T09:08:24.364931Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-07-05T09:08:24.364931Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"q9i9RA6DrNPUWlWPEpxBSmYrLxo6OMAs4FMah1Fki115Kd/wQU13x6PsjdnbePlE2ZQ7Ur48gG1eCmak3XyMDA==","signature_status":"signed_v1","signed_at":"2026-07-05T09:08:24.365457Z","signed_message":"canonical_sha256_bytes"},"source_id":"2404.00978","source_kind":"arxiv","source_version":2}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:88c9c3af2f48602e356681e80d0d8b5cd690dcd4d61b40fe034b568a06bde1a8","sha256:1330bb053044076f900293313f1374e5d4acdd2045fea224f9c47f763226faa0"],"state_sha256":"526ffbde949f4ce86a7d5e534dd38e5af766869bc8314b536d7fc1aac389814e"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"ORhxlOBBb9gGndPizG2WdgzNUP4CANH0mqhoTXFMCNKu9fEzTaa7V1Lq+Ti8NEd/ye4lSkKXf3wNSQ7H3+DIAw==","signed_message":"bundle_sha256_bytes","signed_at":"2026-08-12T05:15:11.057079Z","bundle_sha256":"ddca88450bd29dac6b6d86fc449aa7760ba9f3d90a13d2abd9ffd45a0f8a4bc5"}}