{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2026:QZVCWZPCXSQYQRXNNRUXFWCNGR","merge_version":"pith-open-graph-merge-v1","event_count":2,"valid_event_count":2,"invalid_event_count":0,"equivocation_count":0,"current":{"canonical_record":{"metadata":{"abstract_canon_sha256":"44514237febfe7091fed30834c9849bf363845afb57451f69902fcfce6bf9715","cross_cats_sorted":[],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2026-05-13T13:09:36Z","title_canon_sha256":"649b4779aabb9fccb77f698a33dbc2290f0ffebd8a0bbe469d355409cdfa7fd3"},"schema_version":"1.0","source":{"id":"2605.13486","kind":"arxiv","version":1}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2605.13486","created_at":"2026-05-18T02:44:41Z"},{"alias_kind":"arxiv_version","alias_value":"2605.13486v1","created_at":"2026-05-18T02:44:41Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.13486","created_at":"2026-05-18T02:44:41Z"},{"alias_kind":"pith_short_12","alias_value":"QZVCWZPCXSQY","created_at":"2026-05-18T12:33:37Z"},{"alias_kind":"pith_short_16","alias_value":"QZVCWZPCXSQYQRXN","created_at":"2026-05-18T12:33:37Z"},{"alias_kind":"pith_short_8","alias_value":"QZVCWZPC","created_at":"2026-05-18T12:33:37Z"}],"graph_snapshots":[{"event_id":"sha256:f17a6456a127d83486e12c53d7abd665aa0a4d5322c1f3ae43c079fb51f88663","target":"graph","created_at":"2026-05-18T02:44:41Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"graph_snapshot":{"author_claims":{"count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","strong_count":0},"builder_version":"pith-number-builder-2026-05-17-v1","claims":{"count":4,"items":[{"attestation":"unclaimed","claim_id":"C1","kind":"strongest_claim","source":"verdict.strongest_claim","status":"machine_extracted","text":"R^2-Mem provides a RL-free and low-cost solution for self-improving LLM agents, improving F1 scores by up to 22.6%, while reducing token consumption by 12.9% and search iterations by 20.2%."},{"attestation":"unclaimed","claim_id":"C2","kind":"weakest_assumption","source":"verdict.weakest_assumption","status":"machine_extracted","text":"That the rubric-guided evaluator accurately distinguishes high- and low-quality steps and that the distilled experiences generalize to new queries without overfitting to the offline trajectories."},{"attestation":"unclaimed","claim_id":"C3","kind":"one_line_summary","source":"verdict.one_line_summary","status":"machine_extracted","text":"R^2-Mem distills rubric-scored experiences from high- and low-quality search trajectories to guide LLM agents, raising F1 by up to 22.6% while cutting tokens 12.9% and iterations 20.2%."},{"attestation":"unclaimed","claim_id":"C4","kind":"headline","source":"verdict.pith_extraction.headline","status":"machine_extracted","text":"R^2-Mem distills abstract experiences from scored past trajectories to guide LLM agents away from repeated search errors without reinforcement learning."}],"snapshot_sha256":"e3d6067f3144751a532cc637a305d66483791fbca725fce6d16dd8e62a0e7823"},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"paper":{"abstract_excerpt":"Deep search has recently emerged as a promising paradigm for enabling agents to retrieve fine-grained historical information without heavy memory pre-managed. However, existing deep search agents for memory system repeat past error behaviors because they fail to learn from the prior high- and low-quality search trajectories. To address this limitation, we propose R^2-Mem, a reflective experience framework for memory search systems. In the offline stage, a Rubric-guided Evaluator scores low- and high-quality steps in historical trajectories, and a self-Reflection Learner distills the correspond","authors_text":"Junkang Wu, Wenyu Mao, Xiangnan He, Xiang Wang, Xinyuan Wang","cross_cats":[],"headline":"R^2-Mem distills abstract experiences from scored past trajectories to guide LLM agents away from repeated search errors without reinforcement learning.","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2026-05-13T13:09:36Z","title":"R^2-Mem: Reflective Experience for Memory Search"},"references":{"count":11,"internal_anchors":1,"resolved_work":11,"sample":[{"cited_arxiv_id":"","doi":"10.18653/v1/2024.findings-acl.137","is_internal_anchor":false,"ref_index":1,"title":"M 3-embedding: Multi-linguality, multi-functionality, multi-granularity text embeddings through self-knowledge distillation","work_id":"6f249b44-4b0a-47dd-b5f4-1877fd9269ad","year":2024},{"cited_arxiv_id":"","doi":"10.1162/tacl_a_00023","is_internal_anchor":false,"ref_index":2,"title":"The N arrative QA Reading Comprehension Challenge","work_id":"82006945-c336-4164-9c68-f07125cc331d","year":2026},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":3,"title":"URLhttps://openreview.net/forum?id=jL7fwchScm. William F. Shen, Xinchi Qiu, Chenxi Whitehouse, Lisa Alazraki, Shashwat Goel, Francesco Barbieri, Timon Willi, Akhil Mathur, and Ilias Leontiadis. Rethin","work_id":"549989be-410e-4209-811a-07402cf7e4bc","year":2026},{"cited_arxiv_id":"2508.19828","doi":"","is_internal_anchor":true,"ref_index":4,"title":"Memory-R1: Enhancing Large Language Model Agents to Manage and Utilize Memories via Reinforcement Learning","work_id":"2a070440-2167-4398-be97-c2d4c3ee3541","year":2018},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":5,"title":"Analyze why this planning trace is high-quality or low-quality","work_id":"0f37153a-dcc8-4592-9381-2f0f48dbb2fb","year":null}],"snapshot_sha256":"eb909a04f21ec584e015dfb95e71530fa4e9bc19bd0236adc5c7f0a671f78872"},"source":{"id":"2605.13486","kind":"arxiv","version":1},"verdict":{"created_at":"2026-05-14T19:06:53.510192Z","id":"1cfbb5d1-5aeb-4659-b8a8-451689d5436c","model_set":{"reader":"grok-4.3"},"one_line_summary":"R^2-Mem distills rubric-scored experiences from high- and low-quality search trajectories to guide LLM agents, raising F1 by up to 22.6% while cutting tokens 12.9% and iterations 20.2%.","pipeline_version":"pith-pipeline@v0.9.0","pith_extraction_headline":"R^2-Mem distills abstract experiences from scored past trajectories to guide LLM agents away from repeated search errors without reinforcement learning.","strongest_claim":"R^2-Mem provides a RL-free and low-cost solution for self-improving LLM agents, improving F1 scores by up to 22.6%, while reducing token consumption by 12.9% and search iterations by 20.2%.","weakest_assumption":"That the rubric-guided evaluator accurately distinguishes high- and low-quality steps and that the distilled experiences generalize to new queries without overfitting to the offline trajectories."}},"verdict_id":"1cfbb5d1-5aeb-4659-b8a8-451689d5436c"}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:b7bcc9f47eb7d1ceab714d3bcc01b8594b1a5a95e7289bb319602246036e06b1","target":"record","created_at":"2026-05-18T02:44:41Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"attestation_state":"computed","canonical_record":{"metadata":{"abstract_canon_sha256":"44514237febfe7091fed30834c9849bf363845afb57451f69902fcfce6bf9715","cross_cats_sorted":[],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2026-05-13T13:09:36Z","title_canon_sha256":"649b4779aabb9fccb77f698a33dbc2290f0ffebd8a0bbe469d355409cdfa7fd3"},"schema_version":"1.0","source":{"id":"2605.13486","kind":"arxiv","version":1}},"canonical_sha256":"866a2b65e2bca18846ed6c6972d84d344d8356b8094bb826ed3cdcdbff15f6ac","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"866a2b65e2bca18846ed6c6972d84d344d8356b8094bb826ed3cdcdbff15f6ac","first_computed_at":"2026-05-18T02:44:41.265847Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-05-18T02:44:41.265847Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"/0njiFI62x7OqLHDoxhLC1Fdvr/H/mGUFwtlOfVebteRgbJ80qR72fMpjczGIQSkrCL+0ay9XzfCuMWX9gWfDg==","signature_status":"signed_v1","signed_at":"2026-05-18T02:44:41.266299Z","signed_message":"canonical_sha256_bytes"},"source_id":"2605.13486","source_kind":"arxiv","source_version":1}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:b7bcc9f47eb7d1ceab714d3bcc01b8594b1a5a95e7289bb319602246036e06b1","sha256:f17a6456a127d83486e12c53d7abd665aa0a4d5322c1f3ae43c079fb51f88663"],"state_sha256":"75eed09f82a54fdcefde366ff0c2cd5c067f5173bc6efb5d0b23f9bb5fc2c24d"}