cs.LGNov 2, 2026

One Mistake, Three Layers: Persistence Rules Under Depth Correlation

Authors: Maryam Rahimimovassagh, Michael Brown, Sanjeda Sara Jennifer, Niloofar Yousefi

Organizations: University of Central Florida

Abstract

Probing frozen representations layer by layer is now standard practice, but reporting when a signal first becomes decodable requires a rule for reading a depth curve, and the obvious rule is wrong. On synthetic curves with known ground truth, requiring a probe score to exceed a null-calibrated threshold for three consecutive stages reports an onset nineteen stages early under transformer-like depth correlation, with every resample returning an onset. The failure follows from correlation between neighboring layers, a hazard known in neuroimaging but not, to our knowledge, checked for depth curves. We give an estimator whose error rate is calibrated rather than assumed. Null trajectories are built by permuting labels within each evaluation unit and holding each permutation fixed across depth, preserving the correlation structure the rule must survive; thresholds are per-layer order statistics; and the persistence requirement is calibrated rather than chosen. Applied to pocket information in a frozen co-folding trunk, stratifying 1,820 post-cutoff targets by protein-family and pocket familiarity, the five-stage rule fires on 4 to 7 per cent of null trajectories against a nominal 5 per cent and the three-stage rule on 12 to 17: the synthetic failure reproduces on real data. Pocket membership is decodable above threshold at all 48 blocks in all four regimes, so the onset cannot discriminate them; the pre-declared accumulation depth does, and is latest for familiar families with structurally novel pockets in every feature set. That regime is also the weakest, by a margin surviving controls for length, pocket measure and alignment depth. Separately, a widely used public co-folding checkpoint loads against its own code with 128 trunk parameters missing; those would be randomly initialized, and every representation extracted would be partly noise while appearing normal.

Figures & tables

Appendix figures & tables18 assets

Supplementary material from the paper’s appendix.

Appendix

Explore similar work

Sep 8, 2026

Certified Topological Interaction in Neural Representations: Exact Tests and the Statistic They Require

Class disentanglement--the separation of a representation's class-conditional point clouds along depth and over training--is measured by descriptive curves: the sentence such a study wants to write, layer l+1 is more disentangled than layer l, is an eyeball judgement with no null. We supply the inferential layer for a topological measurement of class overlap, the Intersection Euler Characteristic Profile: the Euler characteristic of the overlap of the clouds' ball unions as a function of scale, from one Alpha-complex sweep with no boundary-matrix reduction. Every number carries a test--exact permutation tests in both directions, a guarded separation certificate the invariant requires, and a paired sign-flip test for comparative claims. Building that test taught a lesson outliving this invariant: its statistic must be scale-free. On the raw profile mass, which has units of feature length, 12,375 paired tests return 5,633 significant steps of which every one at the first epoch points the wrong way, certifying feature-norm dynamics as disentanglement; the dimensionless statistic returns 2,707, with 2,026 decreases. Across 111 networks and 52,650 measurements, disentanglement is depth-graded and early, and interaction quotients rank class pairs by confusability (rho=0.83), on par with cheap separability statistics. In a 96-model factorial, augmentation is the one training choice that separates classes relative to chance; weight decay compresses the overlap without separating. Only a k-fold statistic can pose the structural question: the joint entanglement of a class triple sits below its strongest pair in 97% of triple-layer cells and 99.5% of deep cells, at median ratios far below a measured null floor, in vision encoders and frozen language models--a regularity, not a law. The unnormalized mass predicts test accuracy (R^2=0.94), the quotient does not, and neither beats a linear probe.
Jul 7, 2026

At-Grok Is Not Converged:A Measurement-Validity Audit for Grokking Representation Metrics

On modular arithmetic, a network's embedding keeps compressing for tens of thousands of steps after it has already generalized. Reading effective rank at the grokking transition overstates the converged value by 3-5x on an MLP, and by 1.3-1.5x on a transformer trained to convergence; on the MLP it also erases which cells compress at all. Compression lags the accuracy transition by an amount on the order of the time-to-grok, at least 10,000 steps, rather than coinciding with it. A one-variable ablation shows what sets the lag size: adding LayerNorm to an otherwise identical transformer moves the fraction of compression done by the grok step from 0.87 to 0.25, and a pre-registered control rules out scale invariance as the mechanism. We package this as an audit that separates onset from compression, flags censoring, excludes boundary cells that never fully generalize, and checks that the reference floor has plateaued, with an adversarial suite that caught a false-confidence bug in our own branch. A secondary, MLP-specific depth law linking norm budget to converged floor fails a generality test on a transformer and flips sign under free weight decay. Code and the toolkit are released.
Jun 19, 2026

What Accuracy and Gradient Cosine Miss: Evaluating Feedback Alignment via Scale Stability, Reference Validity, and Depth Utility

Despite the success of deep learning, training deep networks in biologically plausible and hardware-efficient ways remains an open challenge. Feedback alignment (FA) methods address this by replacing backpropagation's symmetric backward weights with fixed random matrices, but their effectiveness depends critically on whether they can be accurately evaluated. The standard evaluation relies on two quantities: task accuracy and cosine similarity between the method's credit signal and the backpropagation gradient. We show that this reporting pair is insufficient by identifying two independent failure modes, both silent under current reporting: (1) measurement degeneracy, where the BP reference gradient collapses to the numerical floor in terminal-LayerNorm residual architectures, rendering cosine uninterpretable; and (2) aggregation collapse, where the aggregate cosine masks layerwise heterogeneity that concentrates credit at one end of the network. To address these limitations, we propose a diagnostic evaluation protocol based on three checks -- scale stability, reference validity, and depth utility -- together with per-layer rather than aggregate cosine reporting. Across multiple architectures and methods, the standard reporting pair gives no signal of failure in any audited case, while our protocol identifies all failures with wide calibration margins. The two failure modes are causally independent: a per-block scale penalty alleviates Mode 1 (residual scale explosion driving reference collapse) without affecting Mode 2 (cosine ranking that contradicts every functional metric we measured). Identifying these silent failures prevents researchers from building on non-functional credit assignment and provides actionable guidance for developing FA methods that genuinely train deep layers.