{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:3L4KJBNPCOBRO244YDL5AGTQB3","short_pith_number":"pith:3L4KJBNP","schema_version":"1.0","canonical_sha256":"daf8a485af1383176b9cc0d7d01a700ef64a8af0696cdc5d98fe937a3a1bf5ba","source":{"kind":"arxiv","id":"2606.05784","version":1},"attestation_state":"computed","paper":{"title":"TAPO: Tool-Aware Policy Optimization via Credit Transfer for Multimodal Search Agents","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.AI","authors_text":"Chengqi Dong, Chuhuai Yue, Fenghe Tang, Guojun Yin, Hang He, Jiajun Chai, S Kevin Zhou, Xiaohan Wang, Yandong Liu","submitted_at":"2026-06-04T07:15:43Z","abstract_excerpt":"We identify and formally characterize credit misassignment as a systematic failure mode of GRPO in tool-augmented multimodal search agents: its uniform broadcast of trajectory-level advantages to all tokens causes valuable tool-use steps in failing trajectories to be penalized no differently from valueless ones. We further empirically quantify the scale of this phenomenon. Over half of failing trajectories and failing tool-use actions exhibit correctable credit misassignment, demonstrating that the wasted training signal is both substantial and structurally exploitable. Building on this insigh"},"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":"2606.05784","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.AI","submitted_at":"2026-06-04T07:15:43Z","cross_cats_sorted":[],"title_canon_sha256":"bce9c072befc312de843e855dc8ee0b7d4884c32219f361155070e1eb00d7ab2","abstract_canon_sha256":"51c2162cb4fa3b0309c8413aa404134c6fb6a585f1d8a5474f3249eb9652e5f3"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-05T01:15:03.107874Z","signature_b64":"1XDb2ZumfBzfTkXScoDaNJV8TSmRkPmO4u3LIRb45C6/kgkozLfJpzkp+igpO3B/9/sG3CU2Eib5LqRKruhbDg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"daf8a485af1383176b9cc0d7d01a700ef64a8af0696cdc5d98fe937a3a1bf5ba","last_reissued_at":"2026-06-05T01:15:03.107453Z","signature_status":"signed_v1","first_computed_at":"2026-06-05T01:15:03.107453Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"TAPO: Tool-Aware Policy Optimization via Credit Transfer for Multimodal Search Agents","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.AI","authors_text":"Chengqi Dong, Chuhuai Yue, Fenghe Tang, Guojun Yin, Hang He, Jiajun Chai, S Kevin Zhou, Xiaohan Wang, Yandong Liu","submitted_at":"2026-06-04T07:15:43Z","abstract_excerpt":"We identify and formally characterize credit misassignment as a systematic failure mode of GRPO in tool-augmented multimodal search agents: its uniform broadcast of trajectory-level advantages to all tokens causes valuable tool-use steps in failing trajectories to be penalized no differently from valueless ones. We further empirically quantify the scale of this phenomenon. Over half of failing trajectories and failing tool-use actions exhibit correctable credit misassignment, demonstrating that the wasted training signal is both substantial and structurally exploitable. Building on this insigh"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2606.05784","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/2606.05784/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":"2606.05784","created_at":"2026-06-05T01:15:03.107511+00:00"},{"alias_kind":"arxiv_version","alias_value":"2606.05784v1","created_at":"2026-06-05T01:15:03.107511+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2606.05784","created_at":"2026-06-05T01:15:03.107511+00:00"},{"alias_kind":"pith_short_12","alias_value":"3L4KJBNPCOBR","created_at":"2026-06-05T01:15:03.107511+00:00"},{"alias_kind":"pith_short_16","alias_value":"3L4KJBNPCOBRO244","created_at":"2026-06-05T01:15:03.107511+00:00"},{"alias_kind":"pith_short_8","alias_value":"3L4KJBNP","created_at":"2026-06-05T01:15:03.107511+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/3L4KJBNPCOBRO244YDL5AGTQB3","json":"https://pith.science/pith/3L4KJBNPCOBRO244YDL5AGTQB3.json","graph_json":"https://pith.science/api/pith-number/3L4KJBNPCOBRO244YDL5AGTQB3/graph.json","events_json":"https://pith.science/api/pith-number/3L4KJBNPCOBRO244YDL5AGTQB3/events.json","paper":"https://pith.science/paper/3L4KJBNP"},"agent_actions":{"view_html":"https://pith.science/pith/3L4KJBNPCOBRO244YDL5AGTQB3","download_json":"https://pith.science/pith/3L4KJBNPCOBRO244YDL5AGTQB3.json","view_paper":"https://pith.science/paper/3L4KJBNP","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2606.05784&json=true","fetch_graph":"https://pith.science/api/pith-number/3L4KJBNPCOBRO244YDL5AGTQB3/graph.json","fetch_events":"https://pith.science/api/pith-number/3L4KJBNPCOBRO244YDL5AGTQB3/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/3L4KJBNPCOBRO244YDL5AGTQB3/action/timestamp_anchor","attest_storage":"https://pith.science/pith/3L4KJBNPCOBRO244YDL5AGTQB3/action/storage_attestation","attest_author":"https://pith.science/pith/3L4KJBNPCOBRO244YDL5AGTQB3/action/author_attestation","sign_citation":"https://pith.science/pith/3L4KJBNPCOBRO244YDL5AGTQB3/action/citation_signature","submit_replication":"https://pith.science/pith/3L4KJBNPCOBRO244YDL5AGTQB3/action/replication_record"}},"created_at":"2026-06-05T01:15:03.107511+00:00","updated_at":"2026-06-05T01:15:03.107511+00:00"}