{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2025:PQELSMEFFT32PBED4PJAX3QAS4","short_pith_number":"pith:PQELSMEF","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"},"canonical_sha256":"7c08b930852cf7a78483e3d20bee00970e581d8d267dd0a8f92ed248170f3299","source":{"kind":"arxiv","id":"2506.15841","version":2},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2506.15841","created_at":"2026-05-17T23:39:19Z"},{"alias_kind":"arxiv_version","alias_value":"2506.15841v2","created_at":"2026-05-17T23:39:19Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2506.15841","created_at":"2026-05-17T23:39:19Z"},{"alias_kind":"pith_short_12","alias_value":"PQELSMEFFT32","created_at":"2026-05-18T12:33:37Z"},{"alias_kind":"pith_short_16","alias_value":"PQELSMEFFT32PBED","created_at":"2026-05-18T12:33:37Z"},{"alias_kind":"pith_short_8","alias_value":"PQELSMEF","created_at":"2026-05-18T12:33:37Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2025:PQELSMEFFT32PBED4PJAX3QAS4","target":"record","payload":{"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"},"canonical_sha256":"7c08b930852cf7a78483e3d20bee00970e581d8d267dd0a8f92ed248170f3299","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"},"source_kind":"arxiv","source_id":"2506.15841","source_version":2,"attestation_state":"computed"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-05-17T23:39:19Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"CSY1UXtNYjUG/jF4mmGIBncuIzOw09V2BetMnWek+4lYh3kAdinKctlTlF7wPKpCt1OHVpOJ7Op+DMiW3n0KCw==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-06-03T16:05:24.192540Z"},"content_sha256":"d7101680098f92b4edbf216e913373d70f8c4ef490d196025c0e2004ba75b2e4","schema_version":"1.0","event_id":"sha256:d7101680098f92b4edbf216e913373d70f8c4ef490d196025c0e2004ba75b2e4"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2025:PQELSMEFFT32PBED4PJAX3QAS4","target":"graph","payload":{"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"},"verdict_id":"8e4d137c-11e8-489a-9a21-fd366fc6660b"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-05-17T23:39:19Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"mnx7/HG/s6YTXtT96xjsfgL9O9MgfMG5lgtcjly5UL0VXdlHusntbwlikhZzGIRJx1lAa98c8bjxzr11VlsaDg==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-06-03T16:05:24.193090Z"},"content_sha256":"c1feb02706b858e78c90e1ac6854c57fcd3de25963ef146a25c61fd9581ccede","schema_version":"1.0","event_id":"sha256:c1feb02706b858e78c90e1ac6854c57fcd3de25963ef146a25c61fd9581ccede"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/PQELSMEFFT32PBED4PJAX3QAS4/bundle.json","state_url":"https://pith.science/pith/PQELSMEFFT32PBED4PJAX3QAS4/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/PQELSMEFFT32PBED4PJAX3QAS4/bundle.json","status":"primary"}],"public_keys":[{"key_id":"pith-v1-2026-05","algorithm":"ed25519","format":"raw","public_key_b64":"stVStoiQhXFxp4s2pdzPNoqVNBMojDU/fJ2db5S3CbM=","public_key_hex":"b2d552b68890857171a78b36a5dccf368a953413288c353f7c9d9d6f94b709b3","fingerprint_sha256_b32_first128bits":"RVFV5Z2OI2J3ZUO7ERDEBCYNKS","fingerprint_sha256_hex":"8d4b5ee74e4693bcd1df2446408b0d54","rotates_at":null,"url":"https://pith.science/pith-signing-key.json","notes":"Pith uses this Ed25519 key to sign canonical record SHA-256 digests. Verify with: ed25519_verify(public_key, message=canonical_sha256_bytes, signature=base64decode(signature_b64))."}],"merge_version":"pith-open-graph-merge-v1","built_at":"2026-06-03T16:05:24Z","links":{"resolver":"https://pith.science/pith/PQELSMEFFT32PBED4PJAX3QAS4","bundle":"https://pith.science/pith/PQELSMEFFT32PBED4PJAX3QAS4/bundle.json","state":"https://pith.science/pith/PQELSMEFFT32PBED4PJAX3QAS4/state.json","well_known_bundle":"https://pith.science/.well-known/pith/PQELSMEFFT32PBED4PJAX3QAS4/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2025:PQELSMEFFT32PBED4PJAX3QAS4","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":"c09ec5a31e75575ba5f80d2b9b9f5dfe7951c0d0433c0a73b84970c066add2e3","cross_cats_sorted":["cs.AI","cs.IR"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2025-06-18T19:44:46Z","title_canon_sha256":"9ac90f2b5b798f1c5e39839f89246cd79b4c1d6500ad89179d71c3d4225e63fd"},"schema_version":"1.0","source":{"id":"2506.15841","kind":"arxiv","version":2}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2506.15841","created_at":"2026-05-17T23:39:19Z"},{"alias_kind":"arxiv_version","alias_value":"2506.15841v2","created_at":"2026-05-17T23:39:19Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2506.15841","created_at":"2026-05-17T23:39:19Z"},{"alias_kind":"pith_short_12","alias_value":"PQELSMEFFT32","created_at":"2026-05-18T12:33:37Z"},{"alias_kind":"pith_short_16","alias_value":"PQELSMEFFT32PBED","created_at":"2026-05-18T12:33:37Z"},{"alias_kind":"pith_short_8","alias_value":"PQELSMEF","created_at":"2026-05-18T12:33:37Z"}],"graph_snapshots":[{"event_id":"sha256:c1feb02706b858e78c90e1ac6854c57fcd3de25963ef146a25c61fd9581ccede","target":"graph","created_at":"2026-05-17T23:39:19Z","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":"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."},{"attestation":"unclaimed","claim_id":"C2","kind":"weakest_assumption","source":"verdict.weakest_assumption","status":"machine_extracted","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."},{"attestation":"unclaimed","claim_id":"C3","kind":"one_line_summary","source":"verdict.one_line_summary","status":"machine_extracted","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."},{"attestation":"unclaimed","claim_id":"C4","kind":"headline","source":"verdict.pith_extraction.headline","status":"machine_extracted","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."}],"snapshot_sha256":"ead35034d430dd1311e76ea6ecbff20594593adf1dcf49f9694339ed010db178"},"formal_canon":{"evidence_count":3,"snapshot_sha256":"ead7ac152b6ecfee3957a83df8244804bbb7da2af6d43009c612d184b18c5a1d"},"paper":{"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 ","authors_text":"Alok Prakash, Ao Qu, Bryan Kian Hsiang Low, Daniela Rus, Jinhua Zhao, Paul Pu Liang, Sunghwan Kim, Zhaoxuan Wu, Zijian Zhou","cross_cats":["cs.AI","cs.IR"],"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.","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2025-06-18T19:44:46Z","title":"MEM1: Learning to Synergize Memory and Reasoning for Efficient Long-Horizon Agents"},"references":{"count":75,"internal_anchors":19,"resolved_work":75,"sample":[{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":1,"title":"Surprising exercises that will sharpen your short- term memory, January 2024","work_id":"dfa227ab-174a-4b6a-8511-4b756b3da24a","year":2024},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":2,"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","year":2025},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":3,"title":"The claude 3 model family: Opus, sonnet, haiku","work_id":"fa4d3ce0-da16-49ff-bf18-86ab716b01b1","year":2024},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":4,"title":"Baddeley and Graham J","work_id":"f6fc23b6-b5aa-4982-9fff-08da2c170dac","year":1974},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":5,"title":"Digirl: Training in-the-wild device-control agents with autonomous reinforcement learning","work_id":"a1622f76-1559-4413-bbd3-5de78aa0af9a","year":2024}],"snapshot_sha256":"ed8a7fa3c5fe3c55c308fd0e3b26e4355b3dbd711ad2925ed9d945ae1106ece0"},"source":{"id":"2506.15841","kind":"arxiv","version":2},"verdict":{"created_at":"2026-05-15T00:22:57.994140Z","id":"8e4d137c-11e8-489a-9a21-fd366fc6660b","model_set":{"reader":"grok-4.3"},"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","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.","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.","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."}},"verdict_id":"8e4d137c-11e8-489a-9a21-fd366fc6660b"}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:d7101680098f92b4edbf216e913373d70f8c4ef490d196025c0e2004ba75b2e4","target":"record","created_at":"2026-05-17T23:39:19Z","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":"c09ec5a31e75575ba5f80d2b9b9f5dfe7951c0d0433c0a73b84970c066add2e3","cross_cats_sorted":["cs.AI","cs.IR"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2025-06-18T19:44:46Z","title_canon_sha256":"9ac90f2b5b798f1c5e39839f89246cd79b4c1d6500ad89179d71c3d4225e63fd"},"schema_version":"1.0","source":{"id":"2506.15841","kind":"arxiv","version":2}},"canonical_sha256":"7c08b930852cf7a78483e3d20bee00970e581d8d267dd0a8f92ed248170f3299","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"7c08b930852cf7a78483e3d20bee00970e581d8d267dd0a8f92ed248170f3299","first_computed_at":"2026-05-17T23:39:19.719794Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-05-17T23:39:19.719794Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"iV3JTLI9xHAVQiLKUwZeeW0IK9piMR9VVyTYfXMeqekes1VQFYB76+BAd9J9YUstH8nrHG0mrxUOIzvuPEfkBQ==","signature_status":"signed_v1","signed_at":"2026-05-17T23:39:19.720522Z","signed_message":"canonical_sha256_bytes"},"source_id":"2506.15841","source_kind":"arxiv","source_version":2}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:d7101680098f92b4edbf216e913373d70f8c4ef490d196025c0e2004ba75b2e4","sha256:c1feb02706b858e78c90e1ac6854c57fcd3de25963ef146a25c61fd9581ccede"],"state_sha256":"4947d34e29ac9153ae2fb6e21906e0dadb89514a3f8810bf75320699ce009c38"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"WLCUAkw/Zv6cvYuCsQ+ej4bkmfNdZQJasGnul5tf9md0yUH1XiXXKwO44ezo3ZBgc67Cw9GdRXjgewJTE/R2Cg==","signed_message":"bundle_sha256_bytes","signed_at":"2026-06-03T16:05:24.195823Z","bundle_sha256":"cd2d89cabd0fece49cf096ea6c1dd762597ddbd49fc818d2a966040c271ec353"}}