{"paper":{"title":"Proactive Memory for Ad-Hoc Recall over Streaming Dialogues","license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","headline":"ProStream maintains a bounded knowledge state for ad-hoc recall over infinite streaming dialogues with higher fidelity and lower latency than baselines.","cross_cats":[],"primary_cat":"cs.AI","authors_text":"Bingbing Wang, Jing Li, Ruifeng Xu","submitted_at":"2026-03-05T07:25:25Z","abstract_excerpt":"Real-world dialogue usually unfolds as an infinite stream. It thus requires bounded-state memory mechanisms to operate within an infinite horizon. However, existing read-then-think memory is fundamentally misaligned with this setting, as it cannot support ad-hoc memory recall while streams unfold. To explore this challenge, we introduce \\textbf{STEM-Bench}, the first benchmark for \\textbf{ST}reaming \\textbf{E}valuation of \\textbf{M}emory. It comprises over 14K QA pairs in dialogue streams that assess perception fidelity, temporal reasoning, and global awareness under infinite-horizon constrain"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"It enables a bounded knowledge state for lower inference latency without sacrificing reasoning fidelity. Experiments show ProStream delivers higher reasoning fidelity than prior baselines while maintaining substantially lower latency than full-context alternatives.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That multi-granular distillation combined with Adaptive Spatiotemporal Optimization can accurately predict retention utility and preserve all necessary information across arbitrary-length streams without introducing critical omissions or errors.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"ProStream achieves higher reasoning fidelity than baselines with substantially lower latency than full-context models by maintaining a bounded knowledge state through hierarchical distillation and adaptive spatiotemporal optimization for ad-hoc recall in streaming dialogues.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"ProStream maintains a bounded knowledge state for ad-hoc recall over infinite streaming dialogues with higher fidelity and lower latency than baselines.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"032e4b0d35f37e22fbc2d5c9b8ee158169e801aacef54eb2f8eaf332fa54e040"},"source":{"id":"2603.04885","kind":"arxiv","version":2},"verdict":{"id":"7b0891f7-b05b-407d-9c5f-bb5d6f408aff","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-15T17:04:42.604975Z","strongest_claim":"It enables a bounded knowledge state for lower inference latency without sacrificing reasoning fidelity. Experiments show ProStream delivers higher reasoning fidelity than prior baselines while maintaining substantially lower latency than full-context alternatives.","one_line_summary":"ProStream achieves higher reasoning fidelity than baselines with substantially lower latency than full-context models by maintaining a bounded knowledge state through hierarchical distillation and adaptive spatiotemporal optimization for ad-hoc recall in streaming dialogues.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That multi-granular distillation combined with Adaptive Spatiotemporal Optimization can accurately predict retention utility and preserve all necessary information across arbitrary-length streams without introducing critical omissions or errors.","pith_extraction_headline":"ProStream maintains a bounded knowledge state for ad-hoc recall over infinite streaming dialogues with higher fidelity and lower latency than baselines."},"references":{"count":0,"sample":[],"resolved_work":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","internal_anchors":0},"formal_canon":{"evidence_count":2,"snapshot_sha256":"a675cbc316d024bc6813c2bc0f4b724cfea4d60afdd1f50556780e37890f4246"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"}