{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:RCDZCAHEWYKFWQ62Q7GT7WBESQ","short_pith_number":"pith:RCDZCAHE","schema_version":"1.0","canonical_sha256":"88879100e4b6145b43da87cd3fd824942b283e6838cbe2e3ad6770c2886eff21","source":{"kind":"arxiv","id":"2410.08193","version":5},"attestation_state":"computed","paper":{"title":"GenARM: Reward Guided Generation with Autoregressive Reward Model for Test-time Alignment","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Alec Koppel, Bang An, Furong Huang, Sicheng Zhu, Sumitra Ganesh, Udari Madhushani Sehwag, Yuancheng Xu","submitted_at":"2024-10-10T17:58:24Z","abstract_excerpt":"Large Language Models (LLMs) exhibit impressive capabilities but require careful alignment with human preferences. Traditional training-time methods finetune LLMs using human preference datasets but incur significant training costs and require repeated training to handle diverse user preferences. Test-time alignment methods address this by using reward models (RMs) to guide frozen LLMs without retraining. However, existing test-time approaches rely on trajectory-level RMs which are designed to evaluate complete responses, making them unsuitable for autoregressive text generation that requires "},"verification_status":{"content_addressed":true,"pith_receipt":true,"author_attested":false,"weak_author_claims":0,"strong_author_claims":0,"externally_anchored":false,"storage_verified":false,"citation_signatures":0,"replication_records":0,"graph_snapshot":true,"references_resolved":false,"formal_links_present":false},"canonical_record":{"source":{"id":"2410.08193","kind":"arxiv","version":5},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2024-10-10T17:58:24Z","cross_cats_sorted":[],"title_canon_sha256":"801f582ffd8dbe1be02009aa5d5de0a93b54b44c492e319f29cd12bc46c3b550","abstract_canon_sha256":"6593968d566dfda3f515354543d091729369d54eef476502a3c574e2bb06b306"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:37:11.462027Z","signature_b64":"qzSseq6yz5naD9V3Y5SMqqOGZMfDLfkiPzUq0jHB8nDXcMOB38BqyzESWF7v9ywYWbeA/jx5epbSN+37KW0tCw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"88879100e4b6145b43da87cd3fd824942b283e6838cbe2e3ad6770c2886eff21","last_reissued_at":"2026-07-05T11:37:11.461594Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:37:11.461594Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"GenARM: Reward Guided Generation with Autoregressive Reward Model for Test-time Alignment","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Alec Koppel, Bang An, Furong Huang, Sicheng Zhu, Sumitra Ganesh, Udari Madhushani Sehwag, Yuancheng Xu","submitted_at":"2024-10-10T17:58:24Z","abstract_excerpt":"Large Language Models (LLMs) exhibit impressive capabilities but require careful alignment with human preferences. Traditional training-time methods finetune LLMs using human preference datasets but incur significant training costs and require repeated training to handle diverse user preferences. Test-time alignment methods address this by using reward models (RMs) to guide frozen LLMs without retraining. However, existing test-time approaches rely on trajectory-level RMs which are designed to evaluate complete responses, making them unsuitable for autoregressive text generation that requires "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2410.08193","kind":"arxiv","version":5},"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/2410.08193/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"},"aliases":[{"alias_kind":"arxiv","alias_value":"2410.08193","created_at":"2026-07-05T11:37:11.461652+00:00"},{"alias_kind":"arxiv_version","alias_value":"2410.08193v5","created_at":"2026-07-05T11:37:11.461652+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2410.08193","created_at":"2026-07-05T11:37:11.461652+00:00"},{"alias_kind":"pith_short_12","alias_value":"RCDZCAHEWYKF","created_at":"2026-07-05T11:37:11.461652+00:00"},{"alias_kind":"pith_short_16","alias_value":"RCDZCAHEWYKFWQ62","created_at":"2026-07-05T11:37:11.461652+00:00"},{"alias_kind":"pith_short_8","alias_value":"RCDZCAHE","created_at":"2026-07-05T11:37:11.461652+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":5,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2505.19075","citing_title":"Universal Reasoner: A Single, Composable Plug-and-Play Reasoner for Frozen LLMs","ref_index":40,"is_internal_anchor":false},{"citing_arxiv_id":"2507.06419","citing_title":"Teach a Reward Model to Correct Itself: Reward Guided Adversarial Failure Discovery for Robust Reward Modeling","ref_index":34,"is_internal_anchor":false},{"citing_arxiv_id":"2605.13875","citing_title":"Common-agency Games for Multi-Objective Test-Time Alignment","ref_index":237,"is_internal_anchor":false},{"citing_arxiv_id":"2605.07353","citing_title":"Confidence-Aware Alignment Makes Reasoning LLMs More Reliable","ref_index":48,"is_internal_anchor":false},{"citing_arxiv_id":"2604.15577","citing_title":"Reward Weighted Classifier-Free Guidance as Policy Improvement in Autoregressive Models","ref_index":15,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/RCDZCAHEWYKFWQ62Q7GT7WBESQ","json":"https://pith.science/pith/RCDZCAHEWYKFWQ62Q7GT7WBESQ.json","graph_json":"https://pith.science/api/pith-number/RCDZCAHEWYKFWQ62Q7GT7WBESQ/graph.json","events_json":"https://pith.science/api/pith-number/RCDZCAHEWYKFWQ62Q7GT7WBESQ/events.json","paper":"https://pith.science/paper/RCDZCAHE"},"agent_actions":{"view_html":"https://pith.science/pith/RCDZCAHEWYKFWQ62Q7GT7WBESQ","download_json":"https://pith.science/pith/RCDZCAHEWYKFWQ62Q7GT7WBESQ.json","view_paper":"https://pith.science/paper/RCDZCAHE","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2410.08193&json=true","fetch_graph":"https://pith.science/api/pith-number/RCDZCAHEWYKFWQ62Q7GT7WBESQ/graph.json","fetch_events":"https://pith.science/api/pith-number/RCDZCAHEWYKFWQ62Q7GT7WBESQ/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/RCDZCAHEWYKFWQ62Q7GT7WBESQ/action/timestamp_anchor","attest_storage":"https://pith.science/pith/RCDZCAHEWYKFWQ62Q7GT7WBESQ/action/storage_attestation","attest_author":"https://pith.science/pith/RCDZCAHEWYKFWQ62Q7GT7WBESQ/action/author_attestation","sign_citation":"https://pith.science/pith/RCDZCAHEWYKFWQ62Q7GT7WBESQ/action/citation_signature","submit_replication":"https://pith.science/pith/RCDZCAHEWYKFWQ62Q7GT7WBESQ/action/replication_record"}},"created_at":"2026-07-05T11:37:11.461652+00:00","updated_at":"2026-07-05T11:37:11.461652+00:00"}