{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:YU3UG4C3YXUAILUPTDCGFH7D7D","short_pith_number":"pith:YU3UG4C3","schema_version":"1.0","canonical_sha256":"c53743705bc5e8042e8f98c4629fe3f8c565342f4f3551bfc285b9f4627fd758","source":{"kind":"arxiv","id":"2508.16153","version":2},"attestation_state":"computed","paper":{"title":"Memento: Fine-tuning LLM Agents without Fine-tuning LLMs","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.LG","authors_text":"Guchun Zhang, Huichi Zhou, Jun Wang, Ka Yiu Lee, Kin Hei Lee, Kun Shao, Linyi Yang, Siyuan Guo, Xue Yan, Yihang Chen, Zihan Wang","submitted_at":"2025-08-22T07:25:30Z","abstract_excerpt":"In this paper, we introduce a novel learning paradigm for Adaptive Large Language Model (LLM) agents that eliminates the need for fine-tuning the underlying LLMs. Existing approaches are often either rigid, relying on static, handcrafted reflection workflows, or computationally intensive, requiring gradient updates of LLM model parameters. In contrast, our method enables low-cost continual adaptation via memory-based online reinforcement learning. We formalise this as a Memory-augmented Markov Decision Process (M-MDP), equipped with a neural case-selection policy to guide action decisions. Pas"},"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":"2508.16153","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-08-22T07:25:30Z","cross_cats_sorted":["cs.CL"],"title_canon_sha256":"a565272a7404cba022f2c35c3dc8f4333a196f1af75350978b59819cd50b1b31","abstract_canon_sha256":"60b3a177ec9ef3d0550b9561cc06b2632a367bedb1f3cbbe00aebb0c9d5346aa"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:58:36.576306Z","signature_b64":"06GBQT71gzVuHyzajw9h0QRjHzVjZP+uUQVe5thXwSvwNHxifgMNePru3Vbf8r+fA27UoB+aUzhsFQDuEur7Ag==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"c53743705bc5e8042e8f98c4629fe3f8c565342f4f3551bfc285b9f4627fd758","last_reissued_at":"2026-07-05T11:58:36.575819Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:58:36.575819Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Memento: Fine-tuning LLM Agents without Fine-tuning LLMs","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.LG","authors_text":"Guchun Zhang, Huichi Zhou, Jun Wang, Ka Yiu Lee, Kin Hei Lee, Kun Shao, Linyi Yang, Siyuan Guo, Xue Yan, Yihang Chen, Zihan Wang","submitted_at":"2025-08-22T07:25:30Z","abstract_excerpt":"In this paper, we introduce a novel learning paradigm for Adaptive Large Language Model (LLM) agents that eliminates the need for fine-tuning the underlying LLMs. Existing approaches are often either rigid, relying on static, handcrafted reflection workflows, or computationally intensive, requiring gradient updates of LLM model parameters. In contrast, our method enables low-cost continual adaptation via memory-based online reinforcement learning. We formalise this as a Memory-augmented Markov Decision Process (M-MDP), equipped with a neural case-selection policy to guide action decisions. Pas"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2508.16153","kind":"arxiv","version":2},"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/2508.16153/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":"2508.16153","created_at":"2026-07-05T11:58:36.575876+00:00"},{"alias_kind":"arxiv_version","alias_value":"2508.16153v2","created_at":"2026-07-05T11:58:36.575876+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2508.16153","created_at":"2026-07-05T11:58:36.575876+00:00"},{"alias_kind":"pith_short_12","alias_value":"YU3UG4C3YXUA","created_at":"2026-07-05T11:58:36.575876+00:00"},{"alias_kind":"pith_short_16","alias_value":"YU3UG4C3YXUAILUP","created_at":"2026-07-05T11:58:36.575876+00:00"},{"alias_kind":"pith_short_8","alias_value":"YU3UG4C3","created_at":"2026-07-05T11:58:36.575876+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":39,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2604.18401","citing_title":"StepPO: Step-Aligned Policy Optimization for Agentic Reinforcement Learning","ref_index":43,"is_internal_anchor":true},{"citing_arxiv_id":"2606.24151","citing_title":"Metis: Bridging Text and Code Memory for Self-Evolving Agents","ref_index":15,"is_internal_anchor":false},{"citing_arxiv_id":"2606.14249","citing_title":"HarnessX: A Composable, Adaptive, and Evolvable Agent Harness Foundry","ref_index":50,"is_internal_anchor":false},{"citing_arxiv_id":"2607.01480","citing_title":"Procedural Memory Distillation: Online Reflection for Self-Improving Language Models","ref_index":55,"is_internal_anchor":false},{"citing_arxiv_id":"2606.11680","citing_title":"Organize then Retrieve: Hierarchical Memory Navigation for Efficient Agents","ref_index":64,"is_internal_anchor":false},{"citing_arxiv_id":"2606.09365","citing_title":"Experience Makes Skillful: Enabling Generalizable Medical Agent Reasoning via Self-Evolving Skill Memory","ref_index":93,"is_internal_anchor":false},{"citing_arxiv_id":"2606.08755","citing_title":"Co-Evolving Skill Generation and Policy Optimization","ref_index":52,"is_internal_anchor":false},{"citing_arxiv_id":"2606.08656","citing_title":"From Player to Master: Enhancing Test-Time Learning of LLM Agents via Reinforcement Learning over Memory","ref_index":21,"is_internal_anchor":false},{"citing_arxiv_id":"2606.07711","citing_title":"Rosetta Memory: Adaptive Memory for Cross-LLM Agents","ref_index":26,"is_internal_anchor":false},{"citing_arxiv_id":"2606.05050","citing_title":"Autonomous heterogeneous catalyst discovery with a self-evolving multi-agent digital twin","ref_index":47,"is_internal_anchor":false},{"citing_arxiv_id":"2606.05250","citing_title":"Towards Persistent Case-Based Memory for Autonomous Data Science: A CBR-Augmented R&D-Agent with a Locally Deployable Small Language Model","ref_index":24,"is_internal_anchor":false},{"citing_arxiv_id":"2606.02812","citing_title":"Traj-Evolve: A Self-Evolving Multi-Agent System for Patient Trajectory Modeling in Lung Cancer Early Detection","ref_index":44,"is_internal_anchor":false},{"citing_arxiv_id":"2606.31121","citing_title":"The Past Is Prologue: A Plug-in Controller for Selective Updates in Sequentially Evolving LLM Memory","ref_index":14,"is_internal_anchor":false},{"citing_arxiv_id":"2605.15102","citing_title":"Improving Multi-turn Dialogue Consistency with Self-Recall Thinking","ref_index":43,"is_internal_anchor":false},{"citing_arxiv_id":"2606.29645","citing_title":"Metadata, Structure, or Strategy? A Decomposition of RAG Context Enrichment","ref_index":32,"is_internal_anchor":false},{"citing_arxiv_id":"2605.28773","citing_title":"Rethinking Memory as Continuously Evolving Connectivity","ref_index":68,"is_internal_anchor":false},{"citing_arxiv_id":"2605.21951","citing_title":"Dynamic Mixture of Latent Memories for Self-Evolving Agents","ref_index":17,"is_internal_anchor":false},{"citing_arxiv_id":"2510.16079","citing_title":"EvolveR: Self-Evolving LLM Agents through an Experience-Driven Lifecycle","ref_index":9,"is_internal_anchor":false},{"citing_arxiv_id":"2605.20189","citing_title":"SOLAR: A Self-Optimizing Open-Ended Autonomous Agent for Lifelong Learning and Continual Adaptation","ref_index":23,"is_internal_anchor":false},{"citing_arxiv_id":"2605.21463","citing_title":"Mem-$\\pi$: Adaptive Memory through Learning When and What to Generate","ref_index":61,"is_internal_anchor":false},{"citing_arxiv_id":"2605.18401","citing_title":"SkillsVote: Lifecycle Governance of Agent Skills from Collection, Recommendation to Evolution","ref_index":79,"is_internal_anchor":false},{"citing_arxiv_id":"2605.18421","citing_title":"EvoMemBench: Benchmarking Agent Memory from a Self-Evolving Perspective","ref_index":45,"is_internal_anchor":false},{"citing_arxiv_id":"2605.19748","citing_title":"Memory-Augmented Reinforcement Learning Agent for CAD Generation","ref_index":85,"is_internal_anchor":false},{"citing_arxiv_id":"2605.15384","citing_title":"Is One Score Enough? Rethinking the Evaluation of Sequentially Evolving LLM Memory","ref_index":47,"is_internal_anchor":false},{"citing_arxiv_id":"2510.16079","citing_title":"EvolveR: Self-Evolving LLM Agents through an Experience-Driven Lifecycle","ref_index":9,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/YU3UG4C3YXUAILUPTDCGFH7D7D","json":"https://pith.science/pith/YU3UG4C3YXUAILUPTDCGFH7D7D.json","graph_json":"https://pith.science/api/pith-number/YU3UG4C3YXUAILUPTDCGFH7D7D/graph.json","events_json":"https://pith.science/api/pith-number/YU3UG4C3YXUAILUPTDCGFH7D7D/events.json","paper":"https://pith.science/paper/YU3UG4C3"},"agent_actions":{"view_html":"https://pith.science/pith/YU3UG4C3YXUAILUPTDCGFH7D7D","download_json":"https://pith.science/pith/YU3UG4C3YXUAILUPTDCGFH7D7D.json","view_paper":"https://pith.science/paper/YU3UG4C3","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2508.16153&json=true","fetch_graph":"https://pith.science/api/pith-number/YU3UG4C3YXUAILUPTDCGFH7D7D/graph.json","fetch_events":"https://pith.science/api/pith-number/YU3UG4C3YXUAILUPTDCGFH7D7D/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/YU3UG4C3YXUAILUPTDCGFH7D7D/action/timestamp_anchor","attest_storage":"https://pith.science/pith/YU3UG4C3YXUAILUPTDCGFH7D7D/action/storage_attestation","attest_author":"https://pith.science/pith/YU3UG4C3YXUAILUPTDCGFH7D7D/action/author_attestation","sign_citation":"https://pith.science/pith/YU3UG4C3YXUAILUPTDCGFH7D7D/action/citation_signature","submit_replication":"https://pith.science/pith/YU3UG4C3YXUAILUPTDCGFH7D7D/action/replication_record"}},"created_at":"2026-07-05T11:58:36.575876+00:00","updated_at":"2026-07-05T11:58:36.575876+00:00"}