{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2025:YABZKNAWN4HUTOI4H4LJ5KBJAS","short_pith_number":"pith:YABZKNAW","canonical_record":{"source":{"id":"2507.09973","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2025-07-14T06:43:00Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"f35af1d25f5891ecf5be09ea081550dc5e0fc632527771b8b7b7807fdb8f51a6","abstract_canon_sha256":"c668121d0383c25fe2283698303dd8bac043038bc6cd1c0e1e7f9a6da6cc7ad1"},"schema_version":"1.0"},"canonical_sha256":"c0039534166f0f49b91c3f169ea82904bae2c86b192ebad0b90011eb4e031b76","source":{"kind":"arxiv","id":"2507.09973","version":1},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2507.09973","created_at":"2026-07-05T11:36:43Z"},{"alias_kind":"arxiv_version","alias_value":"2507.09973v1","created_at":"2026-07-05T11:36:43Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2507.09973","created_at":"2026-07-05T11:36:43Z"},{"alias_kind":"pith_short_12","alias_value":"YABZKNAWN4HU","created_at":"2026-07-05T11:36:43Z"},{"alias_kind":"pith_short_16","alias_value":"YABZKNAWN4HUTOI4","created_at":"2026-07-05T11:36:43Z"},{"alias_kind":"pith_short_8","alias_value":"YABZKNAW","created_at":"2026-07-05T11:36:43Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2025:YABZKNAWN4HUTOI4H4LJ5KBJAS","target":"record","payload":{"canonical_record":{"source":{"id":"2507.09973","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2025-07-14T06:43:00Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"f35af1d25f5891ecf5be09ea081550dc5e0fc632527771b8b7b7807fdb8f51a6","abstract_canon_sha256":"c668121d0383c25fe2283698303dd8bac043038bc6cd1c0e1e7f9a6da6cc7ad1"},"schema_version":"1.0"},"canonical_sha256":"c0039534166f0f49b91c3f169ea82904bae2c86b192ebad0b90011eb4e031b76","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:36:43.082094Z","signature_b64":"BlYu58LTd+UN3MEtTEIL4M65uMzsTKLK0rRLpRehqFsSUFY1C8fOGEfFDcLfWQi1ko4k/J9N3Hw6nciaOHUyDA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"c0039534166f0f49b91c3f169ea82904bae2c86b192ebad0b90011eb4e031b76","last_reissued_at":"2026-07-05T11:36:43.081546Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:36:43.081546Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2507.09973","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":"2026-07-05T11:36:43Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"3X0zZInxLugipqcsOo6rsP5ivrYi8yH+fM1WatFrdVUkIv0GHiJ7OWpbqDHco5hJy0kz0RwJylluGvQIzPlWAw==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-07T18:06:12.654194Z"},"content_sha256":"21e055cab8fba716027b008979ba6ea74950ba07d12a8ab14bdbbf1dc6008f04","schema_version":"1.0","event_id":"sha256:21e055cab8fba716027b008979ba6ea74950ba07d12a8ab14bdbbf1dc6008f04"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2025:YABZKNAWN4HUTOI4H4LJ5KBJAS","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"Tiny Reward Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Sarah Pan","submitted_at":"2025-07-14T06:43:00Z","abstract_excerpt":"Large decoder-based language models have become the dominant architecture for reward modeling in reinforcement learning from human feedback (RLHF). However, as reward models are increasingly deployed in test-time strategies, their inference costs become a growing concern. We present TinyRM, a family of small, bidirectional masked language models (MLMs) with as few as 400 million parameters, that rival the capabilities of models over 175 times larger on reasoning and safety preference modeling tasks. TinyRM combines FLAN-style prompting, Directional Low-Rank Adaptation (DoRA), and layer freezin"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2507.09973","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/2507.09973/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":"2026-07-05T11:36:43Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"lpUYP8izNX03EhDAMLvQDfVXmfanXPp/357kFSOgw1Q/fXxSa47IBCp7tnQKcHTiKgPHp1Ly0bb2/y9RUyEtAw==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-07T18:06:12.654726Z"},"content_sha256":"0b5176c8d48ac95c9335fbc76abdda9910938829a8490225ebc3111f3bb9bff0","schema_version":"1.0","event_id":"sha256:0b5176c8d48ac95c9335fbc76abdda9910938829a8490225ebc3111f3bb9bff0"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/YABZKNAWN4HUTOI4H4LJ5KBJAS/bundle.json","state_url":"https://pith.science/pith/YABZKNAWN4HUTOI4H4LJ5KBJAS/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/YABZKNAWN4HUTOI4H4LJ5KBJAS/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":"2026-08-07T18:06:12Z","links":{"resolver":"https://pith.science/pith/YABZKNAWN4HUTOI4H4LJ5KBJAS","bundle":"https://pith.science/pith/YABZKNAWN4HUTOI4H4LJ5KBJAS/bundle.json","state":"https://pith.science/pith/YABZKNAWN4HUTOI4H4LJ5KBJAS/state.json","well_known_bundle":"https://pith.science/.well-known/pith/YABZKNAWN4HUTOI4H4LJ5KBJAS/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2025:YABZKNAWN4HUTOI4H4LJ5KBJAS","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":"c668121d0383c25fe2283698303dd8bac043038bc6cd1c0e1e7f9a6da6cc7ad1","cross_cats_sorted":["cs.AI"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2025-07-14T06:43:00Z","title_canon_sha256":"f35af1d25f5891ecf5be09ea081550dc5e0fc632527771b8b7b7807fdb8f51a6"},"schema_version":"1.0","source":{"id":"2507.09973","kind":"arxiv","version":1}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2507.09973","created_at":"2026-07-05T11:36:43Z"},{"alias_kind":"arxiv_version","alias_value":"2507.09973v1","created_at":"2026-07-05T11:36:43Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2507.09973","created_at":"2026-07-05T11:36:43Z"},{"alias_kind":"pith_short_12","alias_value":"YABZKNAWN4HU","created_at":"2026-07-05T11:36:43Z"},{"alias_kind":"pith_short_16","alias_value":"YABZKNAWN4HUTOI4","created_at":"2026-07-05T11:36:43Z"},{"alias_kind":"pith_short_8","alias_value":"YABZKNAW","created_at":"2026-07-05T11:36:43Z"}],"graph_snapshots":[{"event_id":"sha256:0b5176c8d48ac95c9335fbc76abdda9910938829a8490225ebc3111f3bb9bff0","target":"graph","created_at":"2026-07-05T11:36:43Z","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/2507.09973/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Large decoder-based language models have become the dominant architecture for reward modeling in reinforcement learning from human feedback (RLHF). However, as reward models are increasingly deployed in test-time strategies, their inference costs become a growing concern. We present TinyRM, a family of small, bidirectional masked language models (MLMs) with as few as 400 million parameters, that rival the capabilities of models over 175 times larger on reasoning and safety preference modeling tasks. TinyRM combines FLAN-style prompting, Directional Low-Rank Adaptation (DoRA), and layer freezin","authors_text":"Sarah Pan","cross_cats":["cs.AI"],"headline":"","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2025-07-14T06:43:00Z","title":"Tiny Reward Models"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2507.09973","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":"sha256:21e055cab8fba716027b008979ba6ea74950ba07d12a8ab14bdbbf1dc6008f04","target":"record","created_at":"2026-07-05T11:36:43Z","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":"c668121d0383c25fe2283698303dd8bac043038bc6cd1c0e1e7f9a6da6cc7ad1","cross_cats_sorted":["cs.AI"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2025-07-14T06:43:00Z","title_canon_sha256":"f35af1d25f5891ecf5be09ea081550dc5e0fc632527771b8b7b7807fdb8f51a6"},"schema_version":"1.0","source":{"id":"2507.09973","kind":"arxiv","version":1}},"canonical_sha256":"c0039534166f0f49b91c3f169ea82904bae2c86b192ebad0b90011eb4e031b76","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"c0039534166f0f49b91c3f169ea82904bae2c86b192ebad0b90011eb4e031b76","first_computed_at":"2026-07-05T11:36:43.081546Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-07-05T11:36:43.081546Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"BlYu58LTd+UN3MEtTEIL4M65uMzsTKLK0rRLpRehqFsSUFY1C8fOGEfFDcLfWQi1ko4k/J9N3Hw6nciaOHUyDA==","signature_status":"signed_v1","signed_at":"2026-07-05T11:36:43.082094Z","signed_message":"canonical_sha256_bytes"},"source_id":"2507.09973","source_kind":"arxiv","source_version":1}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:21e055cab8fba716027b008979ba6ea74950ba07d12a8ab14bdbbf1dc6008f04","sha256:0b5176c8d48ac95c9335fbc76abdda9910938829a8490225ebc3111f3bb9bff0"],"state_sha256":"f055e43b83f4e43b60404a6d31a644b6bf52040426cdbe82f00321b7ac378836"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"kxVoGkEgnbW1pIWsk4n64zV5JjK1h3pMd9zsoHOCAPkryL6sk8PFjiG3GQGt1q2NJdIE2xzGViB7Imls7wyVDw==","signed_message":"bundle_sha256_bytes","signed_at":"2026-08-07T18:06:12.660355Z","bundle_sha256":"95141e18d603a78e948146d863b96a45056d7c9ce43a45faaf6424894851a766"}}