{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2025:BFOX3J775OGVS55N6VDJAQGTER","short_pith_number":"pith:BFOX3J77","canonical_record":{"source":{"id":"2505.02363","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2025-05-05T04:54:44Z","cross_cats_sorted":[],"title_canon_sha256":"1eef84c2a56aed7a9ff8aa2843cf0faa7ecb901a638c9e845890825ff7b2b69f","abstract_canon_sha256":"c76238232c3465c8ec3dc4a4d6d6ab83cf53b33817053818803ed70e6e235585"},"schema_version":"1.0"},"canonical_sha256":"095d7da7ffeb8d5977adf5469040d324464b3bb2293468a18d97efd3a782a534","source":{"kind":"arxiv","id":"2505.02363","version":1},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2505.02363","created_at":"2026-07-05T10:58:35Z"},{"alias_kind":"arxiv_version","alias_value":"2505.02363v1","created_at":"2026-07-05T10:58:35Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2505.02363","created_at":"2026-07-05T10:58:35Z"},{"alias_kind":"pith_short_12","alias_value":"BFOX3J775OGV","created_at":"2026-07-05T10:58:35Z"},{"alias_kind":"pith_short_16","alias_value":"BFOX3J775OGVS55N","created_at":"2026-07-05T10:58:35Z"},{"alias_kind":"pith_short_8","alias_value":"BFOX3J77","created_at":"2026-07-05T10:58:35Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2025:BFOX3J775OGVS55N6VDJAQGTER","target":"record","payload":{"canonical_record":{"source":{"id":"2505.02363","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2025-05-05T04:54:44Z","cross_cats_sorted":[],"title_canon_sha256":"1eef84c2a56aed7a9ff8aa2843cf0faa7ecb901a638c9e845890825ff7b2b69f","abstract_canon_sha256":"c76238232c3465c8ec3dc4a4d6d6ab83cf53b33817053818803ed70e6e235585"},"schema_version":"1.0"},"canonical_sha256":"095d7da7ffeb8d5977adf5469040d324464b3bb2293468a18d97efd3a782a534","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:58:35.941406Z","signature_b64":"oYoxdUT79mPoLcLWwE7sizdNGOZ70/aUBlY2uPmqHxbC5c8wXhLeTz8DpRhcwNMfD4hQHGJBxA1Tv3dlkDQPBQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"095d7da7ffeb8d5977adf5469040d324464b3bb2293468a18d97efd3a782a534","last_reissued_at":"2026-07-05T10:58:35.940999Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:58:35.940999Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2505.02363","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":"2026-07-05T10:58:35Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"WWTvkMiCOR3z0ZwGIG4oqkK03HtUXrhkEVPXsIidxAEm3rUW5Lvwp/2fMO9pzuBcR2O8FYF7y1WZSzfr7EVtCw==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-16T09:10:02.562201Z"},"content_sha256":"6b999a25053ccb909bfa1b91201e8bc02a6f46dfa8541b66b966c66ebe52ca48","schema_version":"1.0","event_id":"sha256:6b999a25053ccb909bfa1b91201e8bc02a6f46dfa8541b66b966c66ebe52ca48"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2025:BFOX3J775OGVS55N6VDJAQGTER","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"SIMPLEMIX: Frustratingly Simple Mixing of Off- and On-policy Data in Language Model Preference Learning","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Daniel Khashabi, Tianjian Li","submitted_at":"2025-05-05T04:54:44Z","abstract_excerpt":"Aligning language models with human preferences relies on pairwise preference datasets. While some studies suggest that on-policy data consistently outperforms off -policy data for preference learning, others indicate that the advantages of on-policy data may be task-dependent, highlighting the need for a systematic exploration of their interplay.\n  In this work, we show that on-policy and off-policy data offer complementary strengths in preference optimization: on-policy data is particularly effective for reasoning tasks like math and coding, while off-policy data performs better on open-ende"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2505.02363","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/2505.02363/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":"2026-07-05T10:58:35Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"s7nlCgGpxKU6CLZyxVXhQWlSbSsBLyvu2+nbHFkHhAEZ2tt292WheLxnrJAN6YUv+YhZ3/feb4pu+RacGgaiCQ==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-16T09:10:02.562694Z"},"content_sha256":"e4dba680c297509f3cde42b369d28e15aa21113a3811837213b5cd0556703c6b","schema_version":"1.0","event_id":"sha256:e4dba680c297509f3cde42b369d28e15aa21113a3811837213b5cd0556703c6b"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/BFOX3J775OGVS55N6VDJAQGTER/bundle.json","state_url":"https://pith.science/pith/BFOX3J775OGVS55N6VDJAQGTER/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/BFOX3J775OGVS55N6VDJAQGTER/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":"2026-08-16T09:10:02Z","links":{"resolver":"https://pith.science/pith/BFOX3J775OGVS55N6VDJAQGTER","bundle":"https://pith.science/pith/BFOX3J775OGVS55N6VDJAQGTER/bundle.json","state":"https://pith.science/pith/BFOX3J775OGVS55N6VDJAQGTER/state.json","well_known_bundle":"https://pith.science/.well-known/pith/BFOX3J775OGVS55N6VDJAQGTER/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2025:BFOX3J775OGVS55N6VDJAQGTER","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":"c76238232c3465c8ec3dc4a4d6d6ab83cf53b33817053818803ed70e6e235585","cross_cats_sorted":[],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2025-05-05T04:54:44Z","title_canon_sha256":"1eef84c2a56aed7a9ff8aa2843cf0faa7ecb901a638c9e845890825ff7b2b69f"},"schema_version":"1.0","source":{"id":"2505.02363","kind":"arxiv","version":1}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2505.02363","created_at":"2026-07-05T10:58:35Z"},{"alias_kind":"arxiv_version","alias_value":"2505.02363v1","created_at":"2026-07-05T10:58:35Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2505.02363","created_at":"2026-07-05T10:58:35Z"},{"alias_kind":"pith_short_12","alias_value":"BFOX3J775OGV","created_at":"2026-07-05T10:58:35Z"},{"alias_kind":"pith_short_16","alias_value":"BFOX3J775OGVS55N","created_at":"2026-07-05T10:58:35Z"},{"alias_kind":"pith_short_8","alias_value":"BFOX3J77","created_at":"2026-07-05T10:58:35Z"}],"graph_snapshots":[{"event_id":"sha256:e4dba680c297509f3cde42b369d28e15aa21113a3811837213b5cd0556703c6b","target":"graph","created_at":"2026-07-05T10:58:35Z","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/2505.02363/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Aligning language models with human preferences relies on pairwise preference datasets. While some studies suggest that on-policy data consistently outperforms off -policy data for preference learning, others indicate that the advantages of on-policy data may be task-dependent, highlighting the need for a systematic exploration of their interplay.\n  In this work, we show that on-policy and off-policy data offer complementary strengths in preference optimization: on-policy data is particularly effective for reasoning tasks like math and coding, while off-policy data performs better on open-ende","authors_text":"Daniel Khashabi, Tianjian Li","cross_cats":[],"headline":"","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2025-05-05T04:54:44Z","title":"SIMPLEMIX: Frustratingly Simple Mixing of Off- and On-policy Data in Language Model Preference Learning"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2505.02363","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":"sha256:6b999a25053ccb909bfa1b91201e8bc02a6f46dfa8541b66b966c66ebe52ca48","target":"record","created_at":"2026-07-05T10:58:35Z","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":"c76238232c3465c8ec3dc4a4d6d6ab83cf53b33817053818803ed70e6e235585","cross_cats_sorted":[],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2025-05-05T04:54:44Z","title_canon_sha256":"1eef84c2a56aed7a9ff8aa2843cf0faa7ecb901a638c9e845890825ff7b2b69f"},"schema_version":"1.0","source":{"id":"2505.02363","kind":"arxiv","version":1}},"canonical_sha256":"095d7da7ffeb8d5977adf5469040d324464b3bb2293468a18d97efd3a782a534","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"095d7da7ffeb8d5977adf5469040d324464b3bb2293468a18d97efd3a782a534","first_computed_at":"2026-07-05T10:58:35.940999Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-07-05T10:58:35.940999Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"oYoxdUT79mPoLcLWwE7sizdNGOZ70/aUBlY2uPmqHxbC5c8wXhLeTz8DpRhcwNMfD4hQHGJBxA1Tv3dlkDQPBQ==","signature_status":"signed_v1","signed_at":"2026-07-05T10:58:35.941406Z","signed_message":"canonical_sha256_bytes"},"source_id":"2505.02363","source_kind":"arxiv","source_version":1}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:6b999a25053ccb909bfa1b91201e8bc02a6f46dfa8541b66b966c66ebe52ca48","sha256:e4dba680c297509f3cde42b369d28e15aa21113a3811837213b5cd0556703c6b"],"state_sha256":"bc248fda78837c5d261389874e6bb826f528ae9afb80496cde2d5f5304a422e9"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"wukMVt+XYAX6us+PB+tPROgPrwtuQqY+y4Nzv2XKB/O9uwLyzJLH7AiZKxsChkmAA0vfpz+OrdXJsHYbIFz4BA==","signed_message":"bundle_sha256_bytes","signed_at":"2026-08-16T09:10:02.567458Z","bundle_sha256":"79d6e46b87fe51ee9de33a7a7cd86eabfdf19c59b2151c36c61c5099a033fa6a"}}