Curriculum-Integrated Depth Scheduling

Mechanism

Curriculum-integrated depth scheduling integrates the curriculum engine with the depth-selective aggregation mechanism to produce a two-dimensional training control framework. The curriculum engine governs the temporal dimension: it determines the order in which training examples from different entropy bands are presented to the model across training epochs. The depth profiles govern the spatial dimension: they determine how deeply each training example's contribution is integrated into the model's layer structure. The curriculum engine controls when the model sees content; the depth profiles control where in the model the content is encoded. The integration of these two dimensions is the regime designated curriculum-integrated depth scheduling.

The depth profile is a structured data object comprising a per-layer or per-block contribution weight vector. For a model comprising L layers or B blocks, the depth profile specifies a weight value for each layer or block, where the weight governs the magnitude of the gradient signal from the associated training example that is permitted to influence the parameters of that layer or block. A weight of one permits the full gradient signal to reach the layer; a weight of zero prevents any gradient signal from reaching the layer; a weight between zero and one attenuates the gradient signal by the specified factor. The per-layer weight vector collectively defines the shape of the training example's contribution across the model's depth dimension.

This scheduling is distinguished from conventional curriculum learning, which orders training examples by difficulty to improve convergence. The present mechanism schedules training examples by semantic governance properties, namely entropy band, policy scope, and provenance, for the purpose of controlled knowledge integration rather than convergence optimization. An example may be deferred not because it is difficult but because its policy scope is not yet authorized for the current training phase, or because its entropy band requires depth profiles that have not yet been activated in the training schedule.

Training Phases

Curriculum-integrated depth scheduling proceeds through defined training phases. In the initial training phase, the curriculum engine presents training examples from all entropy bands with broad exposure, and the depth profiles specify broad, approximately uniform contribution weights across all layer blocks. The objective of the initial phase is to establish foundational representations across the full depth of the model without premature specialization, so that the model develops undifferentiated representations that encode both simple and complex patterns at all depths.

In the intermediate training phase, the curriculum engine begins entropy-band-sequenced presentation, progressively increasing the proportion of mid-entropy and high-entropy content in the training batches. Concurrently the depth profiles begin to narrow: low-entropy content receives contribution weights that increasingly favor shallow blocks, and high-entropy content receives contribution weights that increasingly favor deep blocks. The intermediate phase produces the initial stratification of the model's representations, in which shallow layers begin to specialize in low-entropy pattern encoding and deep layers begin to specialize in high-entropy abstraction.

In the advanced training phase, the curriculum engine presents training batches dominated by high-entropy content, namely complex, novel, semantically dense material that requires deep abstraction to encode. The depth profiles during the advanced phase specify concentrated contribution weights for the deep layer blocks and attenuated weights for the shallow layer blocks. The advanced phase deepens and refines the model's abstract representations while protecting the shallow-layer specialization established during the intermediate phase from disruption by continued gradient flow from high-entropy content.

Adaptive Phase Transition

The transition between training phases is not triggered by a fixed epoch count but by the profile adaptation engine's assessment of the model's internal entropy distribution. The profile adaptation engine evaluates the layer-wise entropy characteristics at defined checkpoints, for example by computing the information-theoretic entropy of the activation distributions at each layer for a held-out evaluation set, and determines whether the model's representations have achieved sufficient stratification to warrant advancement to the next phase. This adaptive phase transition ensures that the training schedule responds to the model's actual learning dynamics rather than to a predetermined timeline.

The same profile adaptation engine adjusts the depth profiles within a phase to maintain alignment between the entropy band structure of the training corpus and the entropy structure of the model's internal representations. If the model's deep layers exhibit low entropy, indicating that deep representations have become overly homogeneous or have failed to develop abstract structure, the engine may increase the deep-layer weights for high-entropy content, directing more complex content toward the underperforming depth range. If the model's shallow layers exhibit high entropy, indicating that shallow representations are overly complex for their intended role as local pattern encoders, the engine may increase the shallow-layer weights for low-entropy content, reinforcing the shallow layers' role as encoders of routine, well-established patterns.

Entropy-Band-Indexed Depth Profiles

Each entropy band recognized by the platform's entropy extraction pipeline is associated with a training depth profile that governs how content from that band is selectively weighted across the layers of the model during training. The entropy band classification is derived from the semantic entropy of each training example, namely the information-theoretic divergence of the example's semantic embedding distribution relative to the model's current representational state, such that examples with low semantic entropy receive shallow depth profiles and examples with high semantic entropy receive deep depth profiles.

The depth profile for high-entropy content specifies elevated contribution weights for the model's deeper layers, where multi-step abstraction, cross-domain integration, and novel pattern synthesis occur. By directing high-entropy content toward deep integration, the system structurally promotes the development of deep representations that encode complex, nuanced knowledge. The depth profile for low-entropy content specifies elevated contribution weights for the model's shallower layers and attenuated or zero contribution weights for the deeper layers, so that routine knowledge does not consume deep representational capacity and the deeper layers are preserved for the complex content that requires multi-step abstraction to encode.

Depth-Selective Aggregation

The depth profiles are applied to the gradient signal during training through depth-selective aggregation, which operates at each layer transition during the backward pass and modulates the gradient signal for each training example according to the depth profile associated with that example's entropy band. The mechanism may be implemented through one or more of three complementary techniques: gated residual connections, in which each residual connection is augmented with a gating coefficient derived from the depth profile; attention-based depth selection, in which the depth-profile weight modulates the gradient flowing through the attention computation during the backward pass; and layer-specific scaling factors, an architecture-agnostic approach that multiplies the gradient signal at each layer boundary by the depth-profile weight before accumulation.

The mechanism commonly operates at block-level granularity rather than at individual-layer granularity. Layers are grouped into blocks, namely contiguous sequences of layers that perform a coherent computational function, and the depth profile specifies per-block aggregation weights. The per-example gradient is scaled by the depth-profile weight at each block before being accumulated into the block's gradient buffer, after the per-example gradient computation and before the batch-level accumulation, so that each example's contribution to each block is individually governed by its depth profile. The mechanism does not alter the optimizer's update rule; it alters the gradient signal the optimizer receives, and is compatible with standard optimization algorithms including stochastic gradient descent, Adam, and AdamW.

Structured Internal Organization

Curriculum-integrated depth scheduling produces models with structured internal knowledge representations organized by semantic complexity. The shallow layers of the trained model encode routine, well-established patterns, namely the low-entropy knowledge that constitutes the model's foundational competence. The intermediate layers encode moderately complex patterns, namely the mid-entropy knowledge that constitutes the model's domain-specific expertise. The deep layers encode highly complex, novel, and abstract patterns, namely the high-entropy knowledge that constitutes the model's capacity for novel reasoning, cross-domain integration, and creative synthesis.

This structured internal organization is an engineered consequence of the two-dimensional control framework that governs both when and where training content is integrated into the model. Because the curriculum controls the order of presentation and the depth profiles control the depth of integration, the resulting stratification is not an emergent accident of optimization but a directed outcome of the governance metadata that accompanies each training example.

Prior-Art Distinction

Conventional curriculum learning orders training examples by difficulty to improve convergence; it does not schedule by entropy band, policy scope, and provenance, and it does not control the depth at which each example's contribution is integrated. Depth-selective aggregation is further distinguished from layer-wise aggregation techniques developed for federated learning, in which different layers receive different aggregation weights across multiple model instances being merged; the present mechanism does not aggregate multiple models but instead governs the depth at which a single training example's gradient contribution is integrated into a single model's parameters, based on the semantic properties of the training content that produced the gradient. The combination disclosed here, in which the curriculum engine governs temporal sequencing while entropy-band-indexed depth profiles govern spatial integration, with phase transitions triggered by the profile adaptation engine's assessment of the model's internal entropy distribution, is the distinguishing feature.

Disclosure Scope

Curriculum-integrated depth scheduling, comprising the two-dimensional control framework in which the curriculum engine governs the temporal sequencing of entropy-banded training content and the depth profiles govern the spatial depth of integration, the initial, intermediate, and advanced training phases, the entropy-band-indexed depth profiles with elevated deep-layer weights for high-entropy content and elevated shallow-layer weights for low-entropy content, the depth-selective aggregation techniques applied during the backward pass at block-level granularity, and the adaptive phase transition triggered by the profile adaptation engine's assessment of the model's internal entropy distribution, is disclosed in the cognition filing (U.S. Application No. 19/647,395 and its international counterpart). This article describes that disclosed mechanism. The scope extends to embodiments employing different gradient-modulation techniques and different block groupings, provided the curriculum sequencing and depth-selective integration remain governed by the platform's semantic metadata.