Adaptive state and Preview
- Adaptive State from Partially Observed Streams explains adaptive state, hybrid representation, and retrieval through Relational Hopfield Memory.
- Introducing Adapt-1 Preview: A Pretraining-Free Substrate for Test-Time Learning introduces the learning system, Domain contracts, retained evidence, and controlled evaluations of learning during operation.
Learning behavior
- The Dynamics of, Sequential, Transitional and Contextual Online Learning follows acquisition and reuse across different learning procedures, including interactive work performed with acquired state frozen.
- Unlocking Plasticity Rules examines temporal context and learned source utility, with controlled evaluations of what each contributes.
Discovery and continuous control
- Emergence: Toward Autonomous Structure Discovery in Adapt-1 describes how Adapt-1 forms learner-facing structure from admitted observations and outcomes.
- Adapt-1 Machina: Learning Continuous Control Sequences follows sequence acquisition from execution outcomes, structural revision, and contextual refinement, with frozen evaluation on unseen conditions.
Value learning and planning
- Coupled Learning and Planning describes native value model training, candidate validation and adoption, continued learning, and planning with learned values.