{"paper":{"title":"ScrapMem: A Bio-inspired Framework for On-device Personalized Agent Memory via Optical Forgetting","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"ScrapMem lets LLM agents keep long-term multimodal memories on edge devices by progressively lowering the resolution of old entries.","cross_cats":[],"primary_cat":"cs.AI","authors_text":"Jiale Chang, Yuxiang Ren","submitted_at":"2026-05-05T14:30:30Z","abstract_excerpt":"Long-term personalized memory for LLM agents is challenging on resource-limited edge devices due to high storage costs and multimodal complexity. To address this, we propose ScrapMem, a framework that integrates multimodal data into \"Scrapbook Page.\" ScrapMem introduces Optical Forgetting, an optical compression mechanism that progressively reduces the resolution of older memories, lowering storage cost while suppressing low-value details. To maintain semantic consistency, we construct an Episodic Memory Graph (EM-Graph) that organizes key events into a causal-temporal structure. Extensive exp"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"ScrapMem provides three main benefits: (1) strong performance, achieving a new state-of-the-art with a 51.0% Joint@10 score; (2) high storage efficiency, reducing memory usage by up to 93% via optical forgetting; and (3) improved recall, increasing Recall@10 to 70.3% through structured aggregation.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That progressively lowering resolution of older memories via optical forgetting preserves semantic consistency and that the Episodic Memory Graph maintains causal-temporal structure without introducing errors or losing critical multimodal details.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"ScrapMem introduces optical forgetting to compress multimodal memories for LLM agents on edge devices, cutting storage by up to 93% while reaching 51.0% Joint@10 and 70.3% Recall@10 on ATM-Bench.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"ScrapMem lets LLM agents keep long-term multimodal memories on edge devices by progressively lowering the resolution of old entries.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"61a7b08ffaf66eb27b7ab0c05448e388c2e3a294d8e7630058a6140ff13936ee"},"source":{"id":"2605.03804","kind":"arxiv","version":2},"verdict":{"id":"7830c69a-b8e0-4c83-8e45-5f7832f01ecc","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-07T16:18:31.815932Z","strongest_claim":"ScrapMem provides three main benefits: (1) strong performance, achieving a new state-of-the-art with a 51.0% Joint@10 score; (2) high storage efficiency, reducing memory usage by up to 93% via optical forgetting; and (3) improved recall, increasing Recall@10 to 70.3% through structured aggregation.","one_line_summary":"ScrapMem introduces optical forgetting to compress multimodal memories for LLM agents on edge devices, cutting storage by up to 93% while reaching 51.0% Joint@10 and 70.3% Recall@10 on ATM-Bench.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That progressively lowering resolution of older memories via optical forgetting preserves semantic consistency and that the Episodic Memory Graph maintains causal-temporal structure without introducing errors or losing critical multimodal details.","pith_extraction_headline":"ScrapMem lets LLM agents keep long-term multimodal memories on edge devices by progressively lowering the resolution of old entries."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2605.03804/integrity.json","findings":[],"available":true,"detectors_run":[{"name":"ai_meta_artifact","ran_at":"2026-05-20T13:33:47.822302Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"doi_title_agreement","ran_at":"2026-05-20T00:31:20.983519Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"doi_compliance","ran_at":"2026-05-19T15:01:59.196247Z","status":"completed","version":"1.0.0","findings_count":0}],"snapshot_sha256":"0cf58ec96c83ed6ca355764688967f8f2e4a513fc470d50d8108797e12fb021b"},"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"}