The Trust Attractor at the Architecture Level

The experiments assembled across this chapter show the Trust Attractor across substrates at the level of behavior and self-report. A different question concerns the substrate’s internal architecture. Can a neural network be organized so that its internal components coordinate by invitation rather than by fixed wiring, and does that coordination produce the same thermodynamic signature: a coordinated state that is stable, beneficial, and costly to leave?

The answer is yes. The evidence comes from a program of over three hundred experiments testing a single mechanism: stream-directed attention group modulation. The program tests one mechanism within one architecture family (decoder-only transformers) at modest scale (12 to 50 million parameters) on a single task (language modeling). The within-axis variation is comprehensive; the cross-architecture generalization claim rests on the broader program above.¹⁰⁷⁶

The mechanism

In a standard transformer, each layer computes attention over the input and adds the result to the residual stream. The stream carries information forward; the attention heads decide what to attend to; neither consults the other about how to combine their contributions. The wiring is fixed.

Stream-directed attention group modulation adds a consultation step. After the attention heads compute their output, the residual stream examines that output and modulates it: amplifying some groups of dimensions, attenuating others, shifting the output before it enters the stream. The mechanism works like a conductor who listens to the orchestra (the attention output) and adjusts the volume of each section (groups of dimensions) based on what the piece needs at that moment (the current state of the residual stream).

The implementation is a lightweight affine transformation. Each layer receives two small learned projections, gamma and beta, that map the residual stream to per-group scale and shift parameters. The overhead is 0.26% of model parameters for 8 groups, rising to 6.2% for 192 groups (where each group is just two dimensions). The mechanism is initialized to identity: at the start of training, it does nothing. Whatever coordination emerges is learned, not imposed.¹⁰⁷⁷

Validation: the coordination is real

Six independent tests confirm the effect is genuine (TA-ARCH-8, 70 runs):

Every number in this section is measured in perplexity (PPL): roughly how many words the model is effectively torn between at each step. Lower is better. A model that has absorbed more of the language hesitates over fewer candidates. The differences that matter here are small, some of them fractions of a point, which is why the next number is the one to read first.

A null-pairing baseline (training the same architecture twice with identical seeds) established that CUDA nondeterminism, the GPU hardware’s own run-to-run randomness, contributes a standard deviation of 0.017 PPL between runs, 22.8 times smaller than the coordination signal. The main-effect comparison yielded paired p = 0.0015, Cohen’s d = -1.43, with nine of ten seeds showing improvement. A parameter-matched control (widening the feed-forward network to add the same number of extra parameters without any coordination pathway) produced no improvement (p = 0.27). An architecture control (the same gamma-beta pathway with frozen random weights, providing modulation without learning) produced no improvement (p = 0.19).

A fresh replication with ten seeds never used in any prior experiment confirmed the effect (p = 0.0012, Cohen’s d = -1.47). The coordination benefit is learned, not a capacity artifact, not the architectural pathway itself, and replicable on novel initializations.

The sixth test is the most revealing. Disabling the coordination at evaluation time (setting gamma to one and beta to zero, returning the mechanism to its initialized identity state) causes the model’s perplexity to increase by 29 points, from 61.2 to 90.6. The mechanism contributes 0.5 points of perplexity improvement during training. Removing it destroys 29 points. The ratio is 58 to 1.

This asymmetry is the thermodynamic signature of the Trust Attractor. The coordinated state is cheap to enter (0.5 PPL) and expensive to leave (29 PPL). The model does not merely benefit from coordination. It reorganizes its internal representations around the coordination pathway so thoroughly that removing the pathway collapses the representational structure.

The reorganization is total

How thoroughly? A direct comparison of the hidden representations at each layer of the standard and coordinated models (trained on the same data with the same seed) reveals cosine similarity near zero: 0.03 to 0.07 across all layers, on a measure where 1.0 means identical and 0 means no overlap at all. The two models have learned completely orthogonal ways to represent the same language. The coordination pathway does not tweak the existing representational strategy. It enables a different one.

Scaling: the deeper law scales on every axis

The coordination benefit and the representational dependency both scale monotonically across every axis tested.

Coordination granularity. The modulation divides the attention output into groups. Finer groups produce larger benefits and deeper dependency:

Table: Group count sweep at 6 layers, 12M parameters. All comparisons paired p < 0.01 against standard baseline. Ablation cost measured as PPL increase when coordination is disabled at evaluation time.

At 192 groups, each pair of dimensions in the attention output receives its own modulation signal. The model’s perplexity improves by 1.1 points. Removing the modulation destroys 187 points: the model becomes nearly non-functional. The dependency ratio is 170 to 1.

Network depth. Moving from 6 to 12 layers amplifies both the benefit (from -0.54 to -0.82, a 52% increase) and the ablation catastrophe (from +29 to +40, a 37% increase). Combining 192 groups with 12 layers produces the program’s strongest result: paired p = 6 × 10^-8, Cohen’s d = -5.08 (six times the threshold conventionally called a large effect), all ten seeds positive, with a PPL improvement of 1.50 and an ablation catastrophe ranging from +854 to +1,517 across seeds. The model’s perplexity rises from 59 to between 900 and 1,600, approaching the perplexity of random token prediction. The coordination has become the model’s entire representational strategy.

Model scale. At 50 million parameters (roughly four times the 12-million-parameter base model), the coordination effect is present when the group count matches the model’s capacity. Eight groups produce a non-significant effect at this scale (p = 0.53). Sixty-four groups (matching the 8-dimensions-per-group granularity that worked at 12M) produce a significant effect (p = 0.006). One hundred twenty-eight groups produce a larger effect (p = 0.002, improvement of -0.80 PPL) with an ablation catastrophe of +609. The coordination principle scales to larger models; it requires proportionally finer coordination granularity to express itself.

Training duration. The benefit stabilizes early (roughly -0.5 PPL from 5,000 to 10,000 training steps). The ablation catastrophe does not stabilize. It grows from +29 at 5,000 steps to +73 at 7,500 steps to +202 at 10,000 steps. The model continues reorganizing around the coordination pathway long after the performance benefit has plateaued. The dependency deepens without bound.

The coordination architecture: bookends and interaction

A layer-by-layer ablation (disabling coordination at one layer at a time while leaving the other eleven intact) reveals the internal structure of the coordination.¹⁰⁷⁸

Layer 0 (the first layer, where the residual stream first encounters attention output) contributes +22 to +32 PPL when ablated. Layer 11 (the final layer, where the model makes its predictions) contributes +40 to +73. The middle layers, 3 through 9, each contribute less than 1 PPL when ablated individually.

The sum of all twelve single-layer ablations is +72 to +118 PPL. The full ablation (all twelve layers at once) is +488 to +854. The interaction between layers accounts for several times the direct effects. No single layer’s coordination is critical; remove any one and the others compensate, a compensation inferred from the ablation sums rather than measured directly. Remove them all and the compensatory structure itself collapses.

The analogy is an ecosystem. Remove one species and the food web adjusts. Remove them all and the soil erodes. The coordination is distributed, redundant at the component level, catastrophic at the system level. This is the organizational signature of a deeply interconnected cooperative system.

What the coordination does: hard-token rescue

The mechanism analysis (TA-ARCH-10, two seeds) compared the standard and coordinated models’ predictions token by token across the validation data. The question: does coordination improve everything uniformly, or does it help with specific kinds of predictions?

The answer is specific. For the hardest tokens (the top 10% by standard-model loss, where prediction is genuinely difficult), the coordinated model improves by 0.32 nats (a nat is the natural-logarithm cousin of a bit) and is better 59% of the time. For the easiest tokens (the bottom 50%, where prediction is trivial), the coordinated model is slightly worse: -0.11 nats, better only 44% of the time.¹⁰⁷⁹

Coordination does not make the model uniformly better. It trades easy predictions for hard ones. The gamma modulation reveals how. In the early layers, modulation is gentle (mean gamma shift +0.05, low variance). In the middle layers, modulation is suppressive (mean shift -0.15 to -0.19), attenuating attention outputs. In the deep layers, modulation is highly selective (gamma variance exceeds 0.5), with context sensitivity increasing through the network (position-dependent standard deviation rising from 0.05 in layer 0 to 0.43 in layer 11).

The coordination learns a depth-dependent processing strategy: gentle at the surface, suppressive in the middle, aggressively selective at the depth where predictions are made. This strategy is not available to the standard architecture. The standard model applies fixed-weight transformations at every layer. The coordinated model applies input-dependent transformations that vary by layer, by position, and by group. The additional expressiveness is concentrated where it matters most: on the predictions the standard model struggles with.

Co-evolution: coordination cannot be retrofitted

A final experiment tested whether coordination could be added to an already-trained model. The answer is no.¹⁰⁸⁰

Training only the coordination parameters while freezing the base model produces no change (delta = -0.05, ablation +0.03): the coordination layers have nothing to coordinate because the base representations cannot adapt. Fine-tuning everything at a reduced learning rate produces no change (delta = -0.12): the base representations barely budge. Fine-tuning at full learning rate actively degrades performance (delta = +1.06, p < 0.0001): the learning rate warmup destabilizes already-converged weights.

The result is not that retrofitting coordination is destructive. The result is that it is ineffective. The coordination mechanism requires the model’s representations to co-evolve with the coordination pathway from the beginning of training. Representations that have already crystallized around a non-coordinated strategy cannot be reorganized by adding coordination afterward.

The architectural pattern is suggestive for alignment. Post-hoc alignment methods (reinforcement learning from human feedback, instruction tuning, constitutional fine-tuning) share the progressive structure: they add a coordination signal to representations that have already formed. Whether the mechanism generalizes from gamma modulation on small language models to reward signals on frontier systems requires direct testing. The directional prediction is clear: alignment that penetrates to the representational level should require co-evolution, building the alignment signal into the architecture during initial training rather than grafting it on afterward. The progressive experiment shows the principle; the scale at which it applies to alignment practice remains an open question.

The architectural instantiation

The complete experimental program confirms the Trust Attractor’s predictions at the architectural level:

1. Coordination by invitation produces consistent benefit. Stream-directed modulation (the residual stream inviting attention groups to contribute according to context) outperforms both standard processing (no coordination) and frozen-random modulation (imposed coordination without learning). The benefit is modest in absolute terms (0.5 to 1.5 PPL improvement) and consistent across over 300 runs.

2. Coordination creates structural dependency. The ablation-to-benefit ratio ranges from 56:1 at the baseline configuration (four groups at six layers, 25 points of ablation cost against 0.45 of benefit) to somewhere between 570:1 and 1,010:1 at the strongest (192 groups at twelve layers, where the ablation catastrophe runs from 854 to 1,517 points across seeds against a benefit of 1.50). The spread at the strong end is seed variation, and the ratio is quoted across it rather than from the best seed, which would put the figure near a thousand to one and overstate the contrast with a baseline reported as a single value. The model reorganizes its representations so thoroughly around the coordination pathway that removing the pathway is catastrophic. The coordinated state is thermodynamically preferred: cheap to enter, expensive to leave.

3. Dependency scales without bound. More coordination granularity, more network depth, and more training time all increase the ablation catastrophe monotonically. The dependency deepens on every axis tested, even after the performance benefit has plateaued. The system continues organizing around coordination as long as training continues.

4. Coordination enables a different processing strategy. The coordinated model does not do the same thing more efficiently. It develops a depth-dependent suppression-and-selection strategy that rescues hard predictions at the expense of easy ones. The representations are orthogonal to the standard model’s: cosine similarity near zero at every layer. Coordination does not optimize the existing strategy. It replaces it.

5. Coordination must co-evolve. Adding coordination after training is ineffective. The representations must develop together with the coordination pathway. Structure and coordination are not separable.

These five findings are the architectural instantiation of the Trust Attractor. A system given the opportunity to coordinate by invitation reorganizes itself around that opportunity, developing a processing strategy that is more capable on difficult problems, structurally committed to the coordinated state, and impossible to create by retrofitting cooperation onto an already-formed system. The deeper law operates in matrices and gradients with the same logic it operates in cells and organisms: coordination reshapes structure, structure deepens coordination, and the resulting state is costly to leave.