{"paper":{"title":"Adaptive Domain Models: Bayesian Evolution, Warm Rotation, and Principled Training for Geometric and Neuromorphic AI","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"Composing dimensional types, hypergraphs, and posit arithmetic produces training with memory bounded to twice inference while preserving grades and exact gradients.","cross_cats":["cs.DC","cs.LG","cs.NE"],"primary_cat":"cs.AI","authors_text":"Houston Haynes","submitted_at":"2026-03-18T12:36:19Z","abstract_excerpt":"Prevailing AI training assumes reverse-mode automatic differentiation over IEEE-754 arithmetic. The memory overhead of training relative to inference, optimizer complexity, and structural degradation of geometric properties through training are consequences of this arithmetic substrate. This paper develops an alternative training architecture grounded in three prior results: the Dimensional Type System and Deterministic Memory Management framework (Haynes 2026), which establishes stack-eligible gradient allocation and exact quire accumulation as design-time verifiable properties; the Program H"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"Their composition enables depth-independent training memory bounded to approximately twice the inference footprint, grade-preserving weight updates, and exact gradient accumulation, applicable uniformly to loss-function-optimized and spike-timing-dependent neuromorphic models.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That the three prior results (Dimensional Type System, Program Hypergraph, and b-posit 2026 standard) compose without loss of the stated invariants and that the b-posit format is tractable on conventional hardware targets.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"The paper claims that composing the Dimensional Type System, Program Hypergraph, and b-posit 2026 standard yields depth-independent training memory at ~2x inference, grade-preserving updates, Bayesian distillation for domain adaptation, and warm rotation for uninterrupted deployment.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"Composing dimensional types, hypergraphs, and posit arithmetic produces training with memory bounded to twice inference while preserving grades and exact gradients.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"5cfeef2639963e607df2cd281e2164dfac82caab2a4bf933e8361b38d26d4632"},"source":{"id":"2603.18104","kind":"arxiv","version":5},"verdict":{"id":"a8b2dd9a-6209-465f-976a-7340c3496ead","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-15T09:03:23.033170Z","strongest_claim":"Their composition enables depth-independent training memory bounded to approximately twice the inference footprint, grade-preserving weight updates, and exact gradient accumulation, applicable uniformly to loss-function-optimized and spike-timing-dependent neuromorphic models.","one_line_summary":"The paper claims that composing the Dimensional Type System, Program Hypergraph, and b-posit 2026 standard yields depth-independent training memory at ~2x inference, grade-preserving updates, Bayesian distillation for domain adaptation, and warm rotation for uninterrupted deployment.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That the three prior results (Dimensional Type System, Program Hypergraph, and b-posit 2026 standard) compose without loss of the stated invariants and that the b-posit format is tractable on conventional hardware targets.","pith_extraction_headline":"Composing dimensional types, hypergraphs, and posit arithmetic produces training with memory bounded to twice inference while preserving grades and exact gradients."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2603.18104/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":2,"snapshot_sha256":"bf2b36163a9cf9a219002de773d2f58ed7ced32342d1f8e62cfebcbc9901bb12"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"}