{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:PQELSMEFFT32PBED4PJAX3QAS4","short_pith_number":"pith:PQELSMEF","schema_version":"1.0","canonical_sha256":"7c08b930852cf7a78483e3d20bee00970e581d8d267dd0a8f92ed248170f3299","source":{"kind":"arxiv","id":"2506.15841","version":2},"attestation_state":"computed","paper":{"title":"MEM1: Learning to Synergize Memory and Reasoning for Efficient Long-Horizon Agents","license":"http://creativecommons.org/licenses/by/4.0/","headline":"MEM1 trains agents to keep constant memory in long multi-turn tasks by updating one shared state that merges memory and reasoning via reinforcement learning.","cross_cats":["cs.AI","cs.IR"],"primary_cat":"cs.CL","authors_text":"Alok Prakash, Ao Qu, Bryan Kian Hsiang Low, Daniela Rus, Jinhua Zhao, Paul Pu Liang, Sunghwan Kim, Zhaoxuan Wu, Zijian Zhou","submitted_at":"2025-06-18T19:44:46Z","abstract_excerpt":"Modern language agents must operate over long-horizon, multi-turn interactions, where they retrieve external information, adapt to observations, and answer interdependent queries. Yet, most LLM systems rely on full-context prompting, appending all past turns regardless of their relevance. This leads to unbounded memory growth, increased computational costs, and degraded reasoning performance on out-of-distribution input lengths. We introduce MEM1, an end-to-end reinforcement learning framework that enables agents to operate with constant memory across long multi-turn tasks. At each turn, MEM1 "},"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":true,"formal_links_present":true},"canonical_record":{"source":{"id":"2506.15841","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2025-06-18T19:44:46Z","cross_cats_sorted":["cs.AI","cs.IR"],"title_canon_sha256":"9ac90f2b5b798f1c5e39839f89246cd79b4c1d6500ad89179d71c3d4225e63fd","abstract_canon_sha256":"c09ec5a31e75575ba5f80d2b9b9f5dfe7951c0d0433c0a73b84970c066add2e3"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-17T23:39:19.720522Z","signature_b64":"iV3JTLI9xHAVQiLKUwZeeW0IK9piMR9VVyTYfXMeqekes1VQFYB76+BAd9J9YUstH8nrHG0mrxUOIzvuPEfkBQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"7c08b930852cf7a78483e3d20bee00970e581d8d267dd0a8f92ed248170f3299","last_reissued_at":"2026-05-17T23:39:19.719794Z","signature_status":"signed_v1","first_computed_at":"2026-05-17T23:39:19.719794Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"MEM1: Learning to Synergize Memory and Reasoning for Efficient Long-Horizon Agents","license":"http://creativecommons.org/licenses/by/4.0/","headline":"MEM1 trains agents to keep constant memory in long multi-turn tasks by updating one shared state that merges memory and reasoning via reinforcement learning.","cross_cats":["cs.AI","cs.IR"],"primary_cat":"cs.CL","authors_text":"Alok Prakash, Ao Qu, Bryan Kian Hsiang Low, Daniela Rus, Jinhua Zhao, Paul Pu Liang, Sunghwan Kim, Zhaoxuan Wu, Zijian Zhou","submitted_at":"2025-06-18T19:44:46Z","abstract_excerpt":"Modern language agents must operate over long-horizon, multi-turn interactions, where they retrieve external information, adapt to observations, and answer interdependent queries. Yet, most LLM systems rely on full-context prompting, appending all past turns regardless of their relevance. This leads to unbounded memory growth, increased computational costs, and degraded reasoning performance on out-of-distribution input lengths. We introduce MEM1, an end-to-end reinforcement learning framework that enables agents to operate with constant memory across long multi-turn tasks. At each turn, MEM1 "},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"MEM1-7B improves performance by 3.5x while reducing memory usage by 3.7x compared to Qwen2.5-14B-Instruct on a 16-objective multi-hop QA task, and generalizes beyond the training horizon.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That reinforcement learning on composed multi-turn environments will produce a memory-update policy that reliably retains all information needed for future interdependent queries while discarding only truly irrelevant content.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"MEM1 uses end-to-end RL to learn constant-memory agents that update a shared state for memory and reasoning, delivering 3.5x better performance and 3.7x lower memory use than larger baselines on long-horizon QA and shopping tasks.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"MEM1 trains agents to keep constant memory in long multi-turn tasks by updating one shared state that merges memory and reasoning via reinforcement learning.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"ead35034d430dd1311e76ea6ecbff20594593adf1dcf49f9694339ed010db178"},"source":{"id":"2506.15841","kind":"arxiv","version":2},"verdict":{"id":"8e4d137c-11e8-489a-9a21-fd366fc6660b","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-15T00:22:57.994140Z","strongest_claim":"MEM1-7B improves performance by 3.5x while reducing memory usage by 3.7x compared to Qwen2.5-14B-Instruct on a 16-objective multi-hop QA task, and generalizes beyond the training horizon.","one_line_summary":"MEM1 uses end-to-end RL to learn constant-memory agents that update a shared state for memory and reasoning, delivering 3.5x better performance and 3.7x lower memory use than larger baselines on long-horizon QA and shopping tasks.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That reinforcement learning on composed multi-turn environments will produce a memory-update policy that reliably retains all information needed for future interdependent queries while discarding only truly irrelevant content.","pith_extraction_headline":"MEM1 trains agents to keep constant memory in long multi-turn tasks by updating one shared state that merges memory and reasoning via reinforcement learning."},"references":{"count":75,"sample":[{"doi":"","year":2024,"title":"Surprising exercises that will sharpen your short- term memory, January 2024","work_id":"dfa227ab-174a-4b6a-8511-4b756b3da24a","ref_index":1,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2025,"title":"Why does the effective context length of llms fall short? In Proceedings of the International Conference on Learning Representations (ICLR), 2025","work_id":"5e744371-ff26-48ad-abd7-f26bbe3559c7","ref_index":2,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2024,"title":"The claude 3 model family: Opus, sonnet, haiku","work_id":"fa4d3ce0-da16-49ff-bf18-86ab716b01b1","ref_index":3,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":1974,"title":"Baddeley and Graham J","work_id":"f6fc23b6-b5aa-4982-9fff-08da2c170dac","ref_index":4,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2024,"title":"Digirl: Training in-the-wild device-control agents with autonomous reinforcement learning","work_id":"a1622f76-1559-4413-bbd3-5de78aa0af9a","ref_index":5,"cited_arxiv_id":"","is_internal_anchor":false}],"resolved_work":75,"snapshot_sha256":"ed8a7fa3c5fe3c55c308fd0e3b26e4355b3dbd711ad2925ed9d945ae1106ece0","internal_anchors":19},"formal_canon":{"evidence_count":3,"snapshot_sha256":"ead7ac152b6ecfee3957a83df8244804bbb7da2af6d43009c612d184b18c5a1d"},"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":"2506.15841","created_at":"2026-05-17T23:39:19.719908+00:00"},{"alias_kind":"arxiv_version","alias_value":"2506.15841v2","created_at":"2026-05-17T23:39:19.719908+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2506.15841","created_at":"2026-05-17T23:39:19.719908+00:00"},{"alias_kind":"pith_short_12","alias_value":"PQELSMEFFT32","created_at":"2026-05-18T12:33:37.589309+00:00"},{"alias_kind":"pith_short_16","alias_value":"PQELSMEFFT32PBED","created_at":"2026-05-18T12:33:37.589309+00:00"},{"alias_kind":"pith_short_8","alias_value":"PQELSMEF","created_at":"2026-05-18T12:33:37.589309+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":30,"internal_anchor_count":30,"sample":[{"citing_arxiv_id":"2605.23668","citing_title":"OnePred: Next-Query Prediction via Recursive Intent Memory in Multi-Turn Conversations","ref_index":26,"is_internal_anchor":true},{"citing_arxiv_id":"2605.20616","citing_title":"Auto-Dreamer: Learning Offline Memory Consolidation for Language Agents","ref_index":46,"is_internal_anchor":true},{"citing_arxiv_id":"2604.27859","citing_title":"Rethinking Agentic Reinforcement Learning In Large Language Models","ref_index":131,"is_internal_anchor":true},{"citing_arxiv_id":"2509.02547","citing_title":"The Landscape of Agentic Reinforcement Learning for LLMs: A Survey","ref_index":138,"is_internal_anchor":true},{"citing_arxiv_id":"2511.02805","citing_title":"MemSearcher: Training LLMs to Reason, Search and Manage Memory via End-to-End Reinforcement Learning","ref_index":42,"is_internal_anchor":true},{"citing_arxiv_id":"2512.10371","citing_title":"AgentProg: Empowering Long-Horizon GUI Agents with Program-Guided Context Management","ref_index":66,"is_internal_anchor":true},{"citing_arxiv_id":"2602.13933","citing_title":"HyMem: Hybrid Memory Architecture with Dynamic Retrieval Scheduling","ref_index":48,"is_internal_anchor":true},{"citing_arxiv_id":"2603.12572","citing_title":"LMEB: Long-horizon Memory Embedding Benchmark","ref_index":42,"is_internal_anchor":true},{"citing_arxiv_id":"2511.20857","citing_title":"Evo-Memory: Benchmarking LLM Agent Test-time Learning with Self-Evolving Memory","ref_index":147,"is_internal_anchor":true},{"citing_arxiv_id":"2605.13037","citing_title":"MAP: A Map-then-Act Paradigm for Long-Horizon Interactive Agent Reasoning","ref_index":49,"is_internal_anchor":true},{"citing_arxiv_id":"2604.02522","citing_title":"Opal: Private Memory for Personal AI","ref_index":287,"is_internal_anchor":true},{"citing_arxiv_id":"2604.03679","citing_title":"LightThinker++: From Reasoning Compression to Memory Management","ref_index":71,"is_internal_anchor":true},{"citing_arxiv_id":"2605.11814","citing_title":"MedMemoryBench: Benchmarking Agent Memory in Personalized Healthcare","ref_index":45,"is_internal_anchor":true},{"citing_arxiv_id":"2605.12493","citing_title":"LongMemEval-V2: Evaluating Long-Term Agent Memory Toward Experienced Colleagues","ref_index":27,"is_internal_anchor":true},{"citing_arxiv_id":"2605.11436","citing_title":"Agent-BRACE: Decoupling Beliefs from Actions in Long-Horizon Tasks via Verbalized State Uncertainty","ref_index":25,"is_internal_anchor":true},{"citing_arxiv_id":"2605.03312","citing_title":"MemFlow: Intent-Driven Memory Orchestration for Small Language Model Agents","ref_index":53,"is_internal_anchor":true},{"citing_arxiv_id":"2604.27859","citing_title":"Rethinking Agentic Reinforcement Learning In Large Language Models","ref_index":131,"is_internal_anchor":true},{"citing_arxiv_id":"2605.08477","citing_title":"Do Agents Need to Plan Step-by-Step? Rethinking Planning Horizon in Data-Centric Tool Calling","ref_index":51,"is_internal_anchor":true},{"citing_arxiv_id":"2604.27859","citing_title":"Rethinking Agentic Reinforcement Learning In Large Language Models","ref_index":131,"is_internal_anchor":true},{"citing_arxiv_id":"2605.05583","citing_title":"Belief Memory: Agent Memory Under Partial Observability","ref_index":21,"is_internal_anchor":true},{"citing_arxiv_id":"2605.00702","citing_title":"Learning How and What to Memorize: Cognition-Inspired Two-Stage Optimization for Evolving Memory","ref_index":190,"is_internal_anchor":true},{"citing_arxiv_id":"2605.04097","citing_title":"CTM-AI: A Blueprint for General AI Inspired by a Model of Consciousness","ref_index":27,"is_internal_anchor":true},{"citing_arxiv_id":"2604.19457","citing_title":"Four-Axis Decision Alignment for Long-Horizon Enterprise AI Agents","ref_index":8,"is_internal_anchor":true},{"citing_arxiv_id":"2604.11462","citing_title":"Escaping the Context Bottleneck: Active Context Curation for LLM Agents via Reinforcement Learning","ref_index":2,"is_internal_anchor":true},{"citing_arxiv_id":"2604.09852","citing_title":"MEMENTO: Teaching LLMs to Manage Their Own Context","ref_index":41,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":3,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/PQELSMEFFT32PBED4PJAX3QAS4","json":"https://pith.science/pith/PQELSMEFFT32PBED4PJAX3QAS4.json","graph_json":"https://pith.science/api/pith-number/PQELSMEFFT32PBED4PJAX3QAS4/graph.json","events_json":"https://pith.science/api/pith-number/PQELSMEFFT32PBED4PJAX3QAS4/events.json","paper":"https://pith.science/paper/PQELSMEF"},"agent_actions":{"view_html":"https://pith.science/pith/PQELSMEFFT32PBED4PJAX3QAS4","download_json":"https://pith.science/pith/PQELSMEFFT32PBED4PJAX3QAS4.json","view_paper":"https://pith.science/paper/PQELSMEF","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2506.15841&json=true","fetch_graph":"https://pith.science/api/pith-number/PQELSMEFFT32PBED4PJAX3QAS4/graph.json","fetch_events":"https://pith.science/api/pith-number/PQELSMEFFT32PBED4PJAX3QAS4/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/PQELSMEFFT32PBED4PJAX3QAS4/action/timestamp_anchor","attest_storage":"https://pith.science/pith/PQELSMEFFT32PBED4PJAX3QAS4/action/storage_attestation","attest_author":"https://pith.science/pith/PQELSMEFFT32PBED4PJAX3QAS4/action/author_attestation","sign_citation":"https://pith.science/pith/PQELSMEFFT32PBED4PJAX3QAS4/action/citation_signature","submit_replication":"https://pith.science/pith/PQELSMEFFT32PBED4PJAX3QAS4/action/replication_record"}},"created_at":"2026-05-17T23:39:19.719908+00:00","updated_at":"2026-05-17T23:39:19.719908+00:00"}