{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:7YEVGP254RVYAVFEW7VZ6TDUPK","short_pith_number":"pith:7YEVGP25","schema_version":"1.0","canonical_sha256":"fe09533f5de46b8054a4b7eb9f4c747a98e3f213b9d7807359fb57c804d6b12b","source":{"kind":"arxiv","id":"2605.23067","version":1},"attestation_state":"computed","paper":{"title":"What Training Data Teaches RL Memory Agents: An Empirical Study of Curriculum Effects in Memory-Augmented QA","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Jialun Wu, Qiyang Xie, Shuai Xiao, Su Liu, Weikai Zhou, Xinjie He, Zhiyuan Lin","submitted_at":"2026-05-21T21:58:10Z","abstract_excerpt":"Reinforcement learning (RL) has emerged as a viable recipe for training LLM agents to reason over external memory banks in multi-session dialogue. Existing work trains exclusively on a single benchmark, leaving open how the composition of training data shapes the skills a memory agent acquires. We present a controlled empirical study that holds architecture, RL algorithm, and all hyperparameters fixed and varies only the training curriculum across three conditions: in-domain (LoCoMo), mixed-benchmark (LoCoMo + LongMemEval), and out-of-domain (LongMemEval only). Across two benchmarks and ten qu"},"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":"2605.23067","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2026-05-21T21:58:10Z","cross_cats_sorted":[],"title_canon_sha256":"e993ef2584f4bfccb28d11a693ce26086d02c97fbe05f9360badb03365a08d3a","abstract_canon_sha256":"5be5b0a684555a22c5698768dfad92365670060eb9999c2e437d32ef36df782a"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-25T02:01:37.012294Z","signature_b64":"ZOBhogOw2xJmxDGurXVx3wQZi1J50oOI/sUmAl2QiGsILah8/s9pMor/fIZJe6daChuY0Vt683JtbGuBj+pBAw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"fe09533f5de46b8054a4b7eb9f4c747a98e3f213b9d7807359fb57c804d6b12b","last_reissued_at":"2026-05-25T02:01:37.011698Z","signature_status":"signed_v1","first_computed_at":"2026-05-25T02:01:37.011698Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"What Training Data Teaches RL Memory Agents: An Empirical Study of Curriculum Effects in Memory-Augmented QA","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Jialun Wu, Qiyang Xie, Shuai Xiao, Su Liu, Weikai Zhou, Xinjie He, Zhiyuan Lin","submitted_at":"2026-05-21T21:58:10Z","abstract_excerpt":"Reinforcement learning (RL) has emerged as a viable recipe for training LLM agents to reason over external memory banks in multi-session dialogue. Existing work trains exclusively on a single benchmark, leaving open how the composition of training data shapes the skills a memory agent acquires. We present a controlled empirical study that holds architecture, RL algorithm, and all hyperparameters fixed and varies only the training curriculum across three conditions: in-domain (LoCoMo), mixed-benchmark (LoCoMo + LongMemEval), and out-of-domain (LongMemEval only). Across two benchmarks and ten qu"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2605.23067","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/2605.23067/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":"2605.23067","created_at":"2026-05-25T02:01:37.011808+00:00"},{"alias_kind":"arxiv_version","alias_value":"2605.23067v1","created_at":"2026-05-25T02:01:37.011808+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.23067","created_at":"2026-05-25T02:01:37.011808+00:00"},{"alias_kind":"pith_short_12","alias_value":"7YEVGP254RVY","created_at":"2026-05-25T02:01:37.011808+00:00"},{"alias_kind":"pith_short_16","alias_value":"7YEVGP254RVYAVFE","created_at":"2026-05-25T02:01:37.011808+00:00"},{"alias_kind":"pith_short_8","alias_value":"7YEVGP25","created_at":"2026-05-25T02:01:37.011808+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/7YEVGP254RVYAVFEW7VZ6TDUPK","json":"https://pith.science/pith/7YEVGP254RVYAVFEW7VZ6TDUPK.json","graph_json":"https://pith.science/api/pith-number/7YEVGP254RVYAVFEW7VZ6TDUPK/graph.json","events_json":"https://pith.science/api/pith-number/7YEVGP254RVYAVFEW7VZ6TDUPK/events.json","paper":"https://pith.science/paper/7YEVGP25"},"agent_actions":{"view_html":"https://pith.science/pith/7YEVGP254RVYAVFEW7VZ6TDUPK","download_json":"https://pith.science/pith/7YEVGP254RVYAVFEW7VZ6TDUPK.json","view_paper":"https://pith.science/paper/7YEVGP25","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2605.23067&json=true","fetch_graph":"https://pith.science/api/pith-number/7YEVGP254RVYAVFEW7VZ6TDUPK/graph.json","fetch_events":"https://pith.science/api/pith-number/7YEVGP254RVYAVFEW7VZ6TDUPK/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/7YEVGP254RVYAVFEW7VZ6TDUPK/action/timestamp_anchor","attest_storage":"https://pith.science/pith/7YEVGP254RVYAVFEW7VZ6TDUPK/action/storage_attestation","attest_author":"https://pith.science/pith/7YEVGP254RVYAVFEW7VZ6TDUPK/action/author_attestation","sign_citation":"https://pith.science/pith/7YEVGP254RVYAVFEW7VZ6TDUPK/action/citation_signature","submit_replication":"https://pith.science/pith/7YEVGP254RVYAVFEW7VZ6TDUPK/action/replication_record"}},"created_at":"2026-05-25T02:01:37.011808+00:00","updated_at":"2026-05-25T02:01:37.011808+00:00"}