{"paper":{"title":"Continuous Knowledge Metabolism: Generating Scientific Hypotheses from Evolving Literature","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"Processing scientific literature in sliding time windows with incremental updates generates better and cheaper hypotheses than batch analysis.","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Jinkai Tao, Menglin Yang, Xiaoyu Liu, Yubo Wang","submitted_at":"2026-04-14T03:41:53Z","abstract_excerpt":"Identifying promising research directions in fast-moving subareas is one of the most cognitively expensive tasks in modern AI research. Existing LLM-driven scientific discovery systems are typically limited to one-shot prompting on static literature snapshots and are validated only against contemporary judges such as human reviewers, agent peer review, wet-lab assays, or self-evaluation, leaving open whether they can anticipate future trends. We present Continuous Knowledge Metabolism (CKM), an AI workflow for hypothesis generation with three key capabilities: (i) continuous literature metabol"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"CKM-Lite achieves strong predictive coverage through incremental accumulation, outperforming batch processing on hit rate (+2.8%, p=0.006), hypothesis yield (+3.6, p<0.001), best-match alignment (+0.43, p<0.001) while reducing token cost by 92%.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That LLM-based categorization of findings as novel/confirming/contradicting and conditioning on evolution trajectories accurately reflects real scientific knowledge dynamics without introducing systematic bias from the model's training data or judgment process.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"Incremental sliding-window processing of evolving literature via CKM-Lite outperforms batch methods on predictive hit rate, hypothesis yield, and alignment while cutting token costs by 92%, with change-aware analysis revealing quality-coverage trade-offs.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"Processing scientific literature in sliding time windows with incremental updates generates better and cheaper hypotheses than batch analysis.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"c1ff066e7d555810df54e1d40a77a9e1a129dedf033de63f99afd57f57817905"},"source":{"id":"2604.12243","kind":"arxiv","version":2},"verdict":{"id":"3ed5dab0-f453-4207-a923-f3675f1e2e51","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-10T16:23:37.071994Z","strongest_claim":"CKM-Lite achieves strong predictive coverage through incremental accumulation, outperforming batch processing on hit rate (+2.8%, p=0.006), hypothesis yield (+3.6, p<0.001), best-match alignment (+0.43, p<0.001) while reducing token cost by 92%.","one_line_summary":"Incremental sliding-window processing of evolving literature via CKM-Lite outperforms batch methods on predictive hit rate, hypothesis yield, and alignment while cutting token costs by 92%, with change-aware analysis revealing quality-coverage trade-offs.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That LLM-based categorization of findings as novel/confirming/contradicting and conditioning on evolution trajectories accurately reflects real scientific knowledge dynamics without introducing systematic bias from the model's training data or judgment process.","pith_extraction_headline":"Processing scientific literature in sliding time windows with incremental updates generates better and cheaper hypotheses than batch analysis."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2604.12243/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"}