Search papers, labs, and topics across Lattice.
This paper introduces SoftModel, a neural architecture designed for continual in-service learning that maintains total plasticity by allowing both parameters and topology to evolve dynamically. By employing a governance mechanism that ensures stability amidst continual changes, the model adapts its structure based on demand, effectively removing the limitations imposed by fixed topologies. Experimental results demonstrate that this approach enables ongoing learning across long task sequences while preserving performance, with the governance mechanism evaluated rigorously against standard continual-learning benchmarks.
Lifelong learning just got a major upgrade: SoftModel's dynamic topology adapts in real-time, breaking free from the constraints of fixed neural architectures.
Today, a neural system is almost always used in two phases -- trained, then deployed -- and in that regime it freezes twice: training ends, and the topology itself was never a degree of freedom. We take the opposite premise as an axiom -- total plasticity: no part of a model, including its structure, is ever frozen -- and derive the governance a lifelong learner then requires. The design's target regime is continual, in-service learning: a long-lived model on a non-stationary stream, whose stability comes from governance rather than immobility and whose capacity follows demand. The result is a growable soft model: an algebra of structural operators (width, hierarchy, composition, input interface, grown cycles, attention heads), each exact at application, budgeted, and audited, with adoption decided solely by a held-out reality gate that treats parametric and structural change uniformly. A complete from-scratch system realizes the whole account; its factory surface is operated end-to-end by a production LLM. Two conclusions follow from the axiom by construction: stability under lifelong change becomes an audit property of the lifecycle, and structure that follows demand removes the silent cap a fixed topology places on later capability where the capacity floor binds. A third is measured: in the worlds where this was measured, the marginal value of new capacity was unobservable before adoption, so workable growth governance took its ex-post form. The same governance extends to evaluative signals, and the core method is evaluated on standard continual-learning benchmarks, where governed growth preserves the ability to keep learning along long task sequences. A pre-registered experimental program adjudicates the mechanism and value claims on the tested problems and reports its failures at full prominence; the map -- positive and negative -- is the contribution.
