{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:R4VTASEBVZYHL6WWOWAU7XAE52","short_pith_number":"pith:R4VTASEB","schema_version":"1.0","canonical_sha256":"8f2b304881ae7075fad675814fdc04eeb216836609611dff3ffa1a9885664de0","source":{"kind":"arxiv","id":"2606.09635","version":1},"attestation_state":"computed","paper":{"title":"Gradient-Guided Reward Optimization for Inference-time Alignment","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Hankun Lin, Ruqi Zhang","submitted_at":"2026-06-08T15:33:13Z","abstract_excerpt":"Ensuring the reliability of Large Language Models (LLMs) under distribution drift requires inference-time adaptation. While inference-time alignment methods such as Best-of-$N$ and rejection sampling are widely used, they frame the task as a sampling-intensive, reward-guided search, leading to two key limitations: their performance is bounded by the base model's generation quality, and their reliance on imperfect reward models makes them vulnerable to reward hacking. To address these challenges, we introduce Gradient-Guided Reward Optimization (GGRO), a lightweight inference-time method that p"},"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":"2606.09635","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2026-06-08T15:33:13Z","cross_cats_sorted":[],"title_canon_sha256":"034b9cde875c54e8dc8d5158890771d5b03ef61c7e530d6c74110c556812c44e","abstract_canon_sha256":"f0c83fce2e668d9c506a530ecb0ea085424a28bdd48bb8525c8c9826185507b7"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-09T02:09:00.302173Z","signature_b64":"WLjiLa+p93uy4safsatO/g+1S+NtNsB3YtNXTev0DjUxUG4u206YWKgGjP0XIrV3OOEvymG3AHs5cSywHlYRCw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"8f2b304881ae7075fad675814fdc04eeb216836609611dff3ffa1a9885664de0","last_reissued_at":"2026-06-09T02:09:00.301264Z","signature_status":"signed_v1","first_computed_at":"2026-06-09T02:09:00.301264Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Gradient-Guided Reward Optimization for Inference-time Alignment","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Hankun Lin, Ruqi Zhang","submitted_at":"2026-06-08T15:33:13Z","abstract_excerpt":"Ensuring the reliability of Large Language Models (LLMs) under distribution drift requires inference-time adaptation. While inference-time alignment methods such as Best-of-$N$ and rejection sampling are widely used, they frame the task as a sampling-intensive, reward-guided search, leading to two key limitations: their performance is bounded by the base model's generation quality, and their reliance on imperfect reward models makes them vulnerable to reward hacking. To address these challenges, we introduce Gradient-Guided Reward Optimization (GGRO), a lightweight inference-time method that p"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2606.09635","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/2606.09635/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":"2606.09635","created_at":"2026-06-09T02:09:00.301420+00:00"},{"alias_kind":"arxiv_version","alias_value":"2606.09635v1","created_at":"2026-06-09T02:09:00.301420+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2606.09635","created_at":"2026-06-09T02:09:00.301420+00:00"},{"alias_kind":"pith_short_12","alias_value":"R4VTASEBVZYH","created_at":"2026-06-09T02:09:00.301420+00:00"},{"alias_kind":"pith_short_16","alias_value":"R4VTASEBVZYHL6WW","created_at":"2026-06-09T02:09:00.301420+00:00"},{"alias_kind":"pith_short_8","alias_value":"R4VTASEB","created_at":"2026-06-09T02:09:00.301420+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/R4VTASEBVZYHL6WWOWAU7XAE52","json":"https://pith.science/pith/R4VTASEBVZYHL6WWOWAU7XAE52.json","graph_json":"https://pith.science/api/pith-number/R4VTASEBVZYHL6WWOWAU7XAE52/graph.json","events_json":"https://pith.science/api/pith-number/R4VTASEBVZYHL6WWOWAU7XAE52/events.json","paper":"https://pith.science/paper/R4VTASEB"},"agent_actions":{"view_html":"https://pith.science/pith/R4VTASEBVZYHL6WWOWAU7XAE52","download_json":"https://pith.science/pith/R4VTASEBVZYHL6WWOWAU7XAE52.json","view_paper":"https://pith.science/paper/R4VTASEB","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2606.09635&json=true","fetch_graph":"https://pith.science/api/pith-number/R4VTASEBVZYHL6WWOWAU7XAE52/graph.json","fetch_events":"https://pith.science/api/pith-number/R4VTASEBVZYHL6WWOWAU7XAE52/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/R4VTASEBVZYHL6WWOWAU7XAE52/action/timestamp_anchor","attest_storage":"https://pith.science/pith/R4VTASEBVZYHL6WWOWAU7XAE52/action/storage_attestation","attest_author":"https://pith.science/pith/R4VTASEBVZYHL6WWOWAU7XAE52/action/author_attestation","sign_citation":"https://pith.science/pith/R4VTASEBVZYHL6WWOWAU7XAE52/action/citation_signature","submit_replication":"https://pith.science/pith/R4VTASEBVZYHL6WWOWAU7XAE52/action/replication_record"}},"created_at":"2026-06-09T02:09:00.301420+00:00","updated_at":"2026-06-09T02:09:00.301420+00:00"}