Tag
Tag pages collect manuscripts that share classification tags or author-provided keywords.
Empirical results for the ASRA Observation Engine on the Original ARC corpus (800 tasks): 100% rule-candidate coverage, ~98% cross-demo common-rule consistency, transform-event distributions, training vs evaluation complexity gradient, and branched-per-demo resolution for 17 exception tasks.
Scientific intelligence increasingly depends on systems that reason from interventions rather than merely fit observations. This review synthesizes conceptual foundations from model-based reinforcement learning, active inference, causal inference, information theory, and perturbation biology into a unified architecture-level view of adaptive scientific reasoning under uncertainty.
Adaptive Scientific Reasoning Architecture (ASRA) applied to decision biology: perturbation–response reasoning, world models, and intervention-centric scientific intelligence. Full text available as PDF (versions 1 and 2).