{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2025:YXNXTX5XH2BT3RTTQGFJRSHSG7","short_pith_number":"pith:YXNXTX5X","canonical_record":{"source":{"id":"2505.18952","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-05-25T02:56:18Z","cross_cats_sorted":[],"title_canon_sha256":"0c808edd1852b0e3619d2d23171e244b1035138c373f21687f5979db14acecae","abstract_canon_sha256":"82d627e51838b9a3df0be355bb659f5c2eab6253b9e5292844da1dc9f05b40f8"},"schema_version":"1.0"},"canonical_sha256":"c5db79dfb73e833dc673818a98c8f237f7fa0ecc417f1305b7550433744acb80","source":{"kind":"arxiv","id":"2505.18952","version":1},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2505.18952","created_at":"2026-07-05T11:09:22Z"},{"alias_kind":"arxiv_version","alias_value":"2505.18952v1","created_at":"2026-07-05T11:09:22Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2505.18952","created_at":"2026-07-05T11:09:22Z"},{"alias_kind":"pith_short_12","alias_value":"YXNXTX5XH2BT","created_at":"2026-07-05T11:09:22Z"},{"alias_kind":"pith_short_16","alias_value":"YXNXTX5XH2BT3RTT","created_at":"2026-07-05T11:09:22Z"},{"alias_kind":"pith_short_8","alias_value":"YXNXTX5X","created_at":"2026-07-05T11:09:22Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2025:YXNXTX5XH2BT3RTTQGFJRSHSG7","target":"record","payload":{"canonical_record":{"source":{"id":"2505.18952","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-05-25T02:56:18Z","cross_cats_sorted":[],"title_canon_sha256":"0c808edd1852b0e3619d2d23171e244b1035138c373f21687f5979db14acecae","abstract_canon_sha256":"82d627e51838b9a3df0be355bb659f5c2eab6253b9e5292844da1dc9f05b40f8"},"schema_version":"1.0"},"canonical_sha256":"c5db79dfb73e833dc673818a98c8f237f7fa0ecc417f1305b7550433744acb80","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:09:22.523086Z","signature_b64":"a1YpjXqCUC66Ji7I9bGN2oN3wchEUKVoTgm3gFwgyShcU9tgmT4Q6vrpu4CsHWbBKLGYFJSBqfSHqijvYRLCDw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"c5db79dfb73e833dc673818a98c8f237f7fa0ecc417f1305b7550433744acb80","last_reissued_at":"2026-07-05T11:09:22.522597Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:09:22.522597Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2505.18952","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":"2026-07-05T11:09:22Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"YEYmZdjrgGZdwFJgfChYcONOe11liDFaPKqFkIDHcsBhVPKCQuBVXIu24s5d6alRyOGYg2qVasKYqcJZO1XvDA==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-08T17:01:59.122935Z"},"content_sha256":"7008c40d6631956fc3ede2e747557e6c50d171d51f9020372c65ea49f6097b1f","schema_version":"1.0","event_id":"sha256:7008c40d6631956fc3ede2e747557e6c50d171d51f9020372c65ea49f6097b1f"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2025:YXNXTX5XH2BT3RTTQGFJRSHSG7","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"Online Knowledge Distillation with Reward Guidance","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Chen Jia","submitted_at":"2025-05-25T02:56:18Z","abstract_excerpt":"This work studies knowledge distillation (KD) for large language models (LLMs) through preference optimization. We propose a reward-guided imitation learning framework for sequential KD, formulating a min-max optimization problem between the policy and reward model (RM) to minimize the performance gap between the student and teacher policies. Specifically, the reward optimization is constrained to achieve near-optimality within a confidence set for preference alignment. For preference data construction, we explore both offline and online preference-based KD. Additionally, we reformulate the RM"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2505.18952","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/2505.18952/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":"2026-07-05T11:09:22Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"mAsBhDp/S3l060ZjVTrPhyvhNmJ3f1czszRuucu1Deg19Li7u0ay0HKrtkwQub7tqudP9385mZcIZ9efUYdgAw==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-08T17:01:59.123480Z"},"content_sha256":"f71ccc12aa36360fe4db47f8911519e61c824dd22a43f620e4cbe6c8b5be6455","schema_version":"1.0","event_id":"sha256:f71ccc12aa36360fe4db47f8911519e61c824dd22a43f620e4cbe6c8b5be6455"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/YXNXTX5XH2BT3RTTQGFJRSHSG7/bundle.json","state_url":"https://pith.science/pith/YXNXTX5XH2BT3RTTQGFJRSHSG7/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/YXNXTX5XH2BT3RTTQGFJRSHSG7/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":"2026-08-08T17:01:59Z","links":{"resolver":"https://pith.science/pith/YXNXTX5XH2BT3RTTQGFJRSHSG7","bundle":"https://pith.science/pith/YXNXTX5XH2BT3RTTQGFJRSHSG7/bundle.json","state":"https://pith.science/pith/YXNXTX5XH2BT3RTTQGFJRSHSG7/state.json","well_known_bundle":"https://pith.science/.well-known/pith/YXNXTX5XH2BT3RTTQGFJRSHSG7/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2025:YXNXTX5XH2BT3RTTQGFJRSHSG7","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":"82d627e51838b9a3df0be355bb659f5c2eab6253b9e5292844da1dc9f05b40f8","cross_cats_sorted":[],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-05-25T02:56:18Z","title_canon_sha256":"0c808edd1852b0e3619d2d23171e244b1035138c373f21687f5979db14acecae"},"schema_version":"1.0","source":{"id":"2505.18952","kind":"arxiv","version":1}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2505.18952","created_at":"2026-07-05T11:09:22Z"},{"alias_kind":"arxiv_version","alias_value":"2505.18952v1","created_at":"2026-07-05T11:09:22Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2505.18952","created_at":"2026-07-05T11:09:22Z"},{"alias_kind":"pith_short_12","alias_value":"YXNXTX5XH2BT","created_at":"2026-07-05T11:09:22Z"},{"alias_kind":"pith_short_16","alias_value":"YXNXTX5XH2BT3RTT","created_at":"2026-07-05T11:09:22Z"},{"alias_kind":"pith_short_8","alias_value":"YXNXTX5X","created_at":"2026-07-05T11:09:22Z"}],"graph_snapshots":[{"event_id":"sha256:f71ccc12aa36360fe4db47f8911519e61c824dd22a43f620e4cbe6c8b5be6455","target":"graph","created_at":"2026-07-05T11:09:22Z","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/2505.18952/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"This work studies knowledge distillation (KD) for large language models (LLMs) through preference optimization. We propose a reward-guided imitation learning framework for sequential KD, formulating a min-max optimization problem between the policy and reward model (RM) to minimize the performance gap between the student and teacher policies. Specifically, the reward optimization is constrained to achieve near-optimality within a confidence set for preference alignment. For preference data construction, we explore both offline and online preference-based KD. Additionally, we reformulate the RM","authors_text":"Chen Jia","cross_cats":[],"headline":"","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-05-25T02:56:18Z","title":"Online Knowledge Distillation with Reward Guidance"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2505.18952","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":"sha256:7008c40d6631956fc3ede2e747557e6c50d171d51f9020372c65ea49f6097b1f","target":"record","created_at":"2026-07-05T11:09:22Z","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":"82d627e51838b9a3df0be355bb659f5c2eab6253b9e5292844da1dc9f05b40f8","cross_cats_sorted":[],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-05-25T02:56:18Z","title_canon_sha256":"0c808edd1852b0e3619d2d23171e244b1035138c373f21687f5979db14acecae"},"schema_version":"1.0","source":{"id":"2505.18952","kind":"arxiv","version":1}},"canonical_sha256":"c5db79dfb73e833dc673818a98c8f237f7fa0ecc417f1305b7550433744acb80","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"c5db79dfb73e833dc673818a98c8f237f7fa0ecc417f1305b7550433744acb80","first_computed_at":"2026-07-05T11:09:22.522597Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-07-05T11:09:22.522597Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"a1YpjXqCUC66Ji7I9bGN2oN3wchEUKVoTgm3gFwgyShcU9tgmT4Q6vrpu4CsHWbBKLGYFJSBqfSHqijvYRLCDw==","signature_status":"signed_v1","signed_at":"2026-07-05T11:09:22.523086Z","signed_message":"canonical_sha256_bytes"},"source_id":"2505.18952","source_kind":"arxiv","source_version":1}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:7008c40d6631956fc3ede2e747557e6c50d171d51f9020372c65ea49f6097b1f","sha256:f71ccc12aa36360fe4db47f8911519e61c824dd22a43f620e4cbe6c8b5be6455"],"state_sha256":"9b3601ba6646dd106e5deaf15f8c1d6fce6969ea4dff7d37925b5147c00cecde"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"y+HUP1uZFT2PI5YJnTxDxNV3fs9TjwdkeeOii95YZOSayTHn8NqzipU7L9GOl6iXLQX5CUJJnXqxwLTIT7I1CQ==","signed_message":"bundle_sha256_bytes","signed_at":"2026-08-08T17:01:59.127465Z","bundle_sha256":"5e5fbd61d2c4da424bc22574bbd71567eb2e4516c9ada8d2ce91f26f941986d7"}}