{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:6F4AQ4XYZDQH3IUVKLFH74MP6E","short_pith_number":"pith:6F4AQ4XY","schema_version":"1.0","canonical_sha256":"f1780872f8c8e07da29552ca7ff18ff12ad772c4eb5151c1a4ee8e595e8b5051","source":{"kind":"arxiv","id":"2607.10738","version":1},"attestation_state":"computed","paper":{"title":"To Answer or to Abstain: Mitigating Search-Agent Hallucinations via Abstention-Aware Reinforcement Learning","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.CL"],"primary_cat":"cs.LG","authors_text":"Bei Chen, Chengen Huang, Fengji Zhang, Jacky Keung, Tianyu Fan, Xinyao Niu, Yuxiang Zheng","submitted_at":"2026-07-12T12:43:12Z","abstract_excerpt":"Recent advances in equipping Large Language Models (LLMs) with search tools and outcome-reward reinforcement learning (RL) have achieved new state-of-the-art results on open-domain QA tasks. However, we argue that current training paradigms harbor a critical vulnerability: they predominantly reward correct answers but fail to penalize fabricated ones when retrieval fails, thereby implicitly exacerbating hallucinations. To address this, we propose Abstention-Aware Reinforcement Learning (AWA-RL), which dynamically shapes the abstention reward utilizing the model's query-specific prior capabilit"},"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":"2607.10738","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-07-12T12:43:12Z","cross_cats_sorted":["cs.AI","cs.CL"],"title_canon_sha256":"6876da6031e80bd6c93f71818740378fa8c12f160768066462af6cbc8aefb55b","abstract_canon_sha256":"25d09316b50538233ca0b5df70bdc2290fdab185ed217289d9357378ab8bea16"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-14T01:21:37.544222Z","signature_b64":"wbnIH3MxzHTpJgj6bLreqfuJjmzOBFImLvxW5wNpxjPlor/sIsDSPeH95v94zYliW8AOBpVNOUkqDD+KwMXwCw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"f1780872f8c8e07da29552ca7ff18ff12ad772c4eb5151c1a4ee8e595e8b5051","last_reissued_at":"2026-07-14T01:21:37.543395Z","signature_status":"signed_v1","first_computed_at":"2026-07-14T01:21:37.543395Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"To Answer or to Abstain: Mitigating Search-Agent Hallucinations via Abstention-Aware Reinforcement Learning","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.CL"],"primary_cat":"cs.LG","authors_text":"Bei Chen, Chengen Huang, Fengji Zhang, Jacky Keung, Tianyu Fan, Xinyao Niu, Yuxiang Zheng","submitted_at":"2026-07-12T12:43:12Z","abstract_excerpt":"Recent advances in equipping Large Language Models (LLMs) with search tools and outcome-reward reinforcement learning (RL) have achieved new state-of-the-art results on open-domain QA tasks. However, we argue that current training paradigms harbor a critical vulnerability: they predominantly reward correct answers but fail to penalize fabricated ones when retrieval fails, thereby implicitly exacerbating hallucinations. To address this, we propose Abstention-Aware Reinforcement Learning (AWA-RL), which dynamically shapes the abstention reward utilizing the model's query-specific prior capabilit"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2607.10738","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/2607.10738/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":"2607.10738","created_at":"2026-07-14T01:21:37.543820+00:00"},{"alias_kind":"arxiv_version","alias_value":"2607.10738v1","created_at":"2026-07-14T01:21:37.543820+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2607.10738","created_at":"2026-07-14T01:21:37.543820+00:00"},{"alias_kind":"pith_short_12","alias_value":"6F4AQ4XYZDQH","created_at":"2026-07-14T01:21:37.543820+00:00"},{"alias_kind":"pith_short_16","alias_value":"6F4AQ4XYZDQH3IUV","created_at":"2026-07-14T01:21:37.543820+00:00"},{"alias_kind":"pith_short_8","alias_value":"6F4AQ4XY","created_at":"2026-07-14T01:21:37.543820+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/6F4AQ4XYZDQH3IUVKLFH74MP6E","json":"https://pith.science/pith/6F4AQ4XYZDQH3IUVKLFH74MP6E.json","graph_json":"https://pith.science/api/pith-number/6F4AQ4XYZDQH3IUVKLFH74MP6E/graph.json","events_json":"https://pith.science/api/pith-number/6F4AQ4XYZDQH3IUVKLFH74MP6E/events.json","paper":"https://pith.science/paper/6F4AQ4XY"},"agent_actions":{"view_html":"https://pith.science/pith/6F4AQ4XYZDQH3IUVKLFH74MP6E","download_json":"https://pith.science/pith/6F4AQ4XYZDQH3IUVKLFH74MP6E.json","view_paper":"https://pith.science/paper/6F4AQ4XY","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2607.10738&json=true","fetch_graph":"https://pith.science/api/pith-number/6F4AQ4XYZDQH3IUVKLFH74MP6E/graph.json","fetch_events":"https://pith.science/api/pith-number/6F4AQ4XYZDQH3IUVKLFH74MP6E/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/6F4AQ4XYZDQH3IUVKLFH74MP6E/action/timestamp_anchor","attest_storage":"https://pith.science/pith/6F4AQ4XYZDQH3IUVKLFH74MP6E/action/storage_attestation","attest_author":"https://pith.science/pith/6F4AQ4XYZDQH3IUVKLFH74MP6E/action/author_attestation","sign_citation":"https://pith.science/pith/6F4AQ4XYZDQH3IUVKLFH74MP6E/action/citation_signature","submit_replication":"https://pith.science/pith/6F4AQ4XYZDQH3IUVKLFH74MP6E/action/replication_record"}},"created_at":"2026-07-14T01:21:37.543820+00:00","updated_at":"2026-07-14T01:21:37.543820+00:00"}