{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:C7H663GZTRUDSUUY275Y6QGJCS","short_pith_number":"pith:C7H663GZ","schema_version":"1.0","canonical_sha256":"17cfef6cd99c68395298d7fb8f40c914becf80d1c8b2cff68b117bdd7520bfea","source":{"kind":"arxiv","id":"2605.17877","version":1},"attestation_state":"computed","paper":{"title":"PAIR: Prefix-Aware Internal Reward Model for Multi-Turn Agent Optimization","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.AI","authors_text":"Chanyoung Park, Dongha Lee, Sangwu Park, Wonjoong Kim, Yeonjun In","submitted_at":"2026-05-18T05:39:30Z","abstract_excerpt":"A significant hurdle for current LLMs is the execution of complex, multi-stage tasks. Group Relative Policy Optimization (GRPO) has been emerging as a leading choice, but its reliance on sparse outcome rewards severely limits credit assignment across intermediate steps. Existing remedies such as running full rollouts to assign step-level advantages, calling external LLM judges at each step, or computing intrinsic rewards that require ground-truth answers at every evaluation introduce significant costs or practical constraints. We hypothesize that internal correctness probing over LLM hidden st"},"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":"2605.17877","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.AI","submitted_at":"2026-05-18T05:39:30Z","cross_cats_sorted":[],"title_canon_sha256":"2cb687661584b4670bda496326726501cc84527f28acccdf6bd8859634d0a662","abstract_canon_sha256":"156b5e80282102886002fbbae935440a70302b416bdcc92383bc23b7ffa12486"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-20T00:05:03.086388Z","signature_b64":"WU++YNqSeredFAVVOS1L1JkCW6LPGW0I+02rYk4GFBcheXeuW10A0/bMZDLxR2OQCqLtp3ZXgNlpJohwGo9SCA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"17cfef6cd99c68395298d7fb8f40c914becf80d1c8b2cff68b117bdd7520bfea","last_reissued_at":"2026-05-20T00:05:03.085481Z","signature_status":"signed_v1","first_computed_at":"2026-05-20T00:05:03.085481Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"PAIR: Prefix-Aware Internal Reward Model for Multi-Turn Agent Optimization","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.AI","authors_text":"Chanyoung Park, Dongha Lee, Sangwu Park, Wonjoong Kim, Yeonjun In","submitted_at":"2026-05-18T05:39:30Z","abstract_excerpt":"A significant hurdle for current LLMs is the execution of complex, multi-stage tasks. Group Relative Policy Optimization (GRPO) has been emerging as a leading choice, but its reliance on sparse outcome rewards severely limits credit assignment across intermediate steps. Existing remedies such as running full rollouts to assign step-level advantages, calling external LLM judges at each step, or computing intrinsic rewards that require ground-truth answers at every evaluation introduce significant costs or practical constraints. We hypothesize that internal correctness probing over LLM hidden st"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2605.17877","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/2605.17877/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":"2605.17877","created_at":"2026-05-20T00:05:03.085636+00:00"},{"alias_kind":"arxiv_version","alias_value":"2605.17877v1","created_at":"2026-05-20T00:05:03.085636+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.17877","created_at":"2026-05-20T00:05:03.085636+00:00"},{"alias_kind":"pith_short_12","alias_value":"C7H663GZTRUD","created_at":"2026-05-20T00:05:03.085636+00:00"},{"alias_kind":"pith_short_16","alias_value":"C7H663GZTRUDSUUY","created_at":"2026-05-20T00:05:03.085636+00:00"},{"alias_kind":"pith_short_8","alias_value":"C7H663GZ","created_at":"2026-05-20T00:05:03.085636+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/C7H663GZTRUDSUUY275Y6QGJCS","json":"https://pith.science/pith/C7H663GZTRUDSUUY275Y6QGJCS.json","graph_json":"https://pith.science/api/pith-number/C7H663GZTRUDSUUY275Y6QGJCS/graph.json","events_json":"https://pith.science/api/pith-number/C7H663GZTRUDSUUY275Y6QGJCS/events.json","paper":"https://pith.science/paper/C7H663GZ"},"agent_actions":{"view_html":"https://pith.science/pith/C7H663GZTRUDSUUY275Y6QGJCS","download_json":"https://pith.science/pith/C7H663GZTRUDSUUY275Y6QGJCS.json","view_paper":"https://pith.science/paper/C7H663GZ","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2605.17877&json=true","fetch_graph":"https://pith.science/api/pith-number/C7H663GZTRUDSUUY275Y6QGJCS/graph.json","fetch_events":"https://pith.science/api/pith-number/C7H663GZTRUDSUUY275Y6QGJCS/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/C7H663GZTRUDSUUY275Y6QGJCS/action/timestamp_anchor","attest_storage":"https://pith.science/pith/C7H663GZTRUDSUUY275Y6QGJCS/action/storage_attestation","attest_author":"https://pith.science/pith/C7H663GZTRUDSUUY275Y6QGJCS/action/author_attestation","sign_citation":"https://pith.science/pith/C7H663GZTRUDSUUY275Y6QGJCS/action/citation_signature","submit_replication":"https://pith.science/pith/C7H663GZTRUDSUUY275Y6QGJCS/action/replication_record"}},"created_at":"2026-05-20T00:05:03.085636+00:00","updated_at":"2026-05-20T00:05:03.085636+00:00"}