arXiv subjects

Adil Amin

Publications and source records attributed to Adil Amin.

6 recordsLinked to original sources

arXiv2026

Lying Is Just a Phase: The Hidden Alignment Transition in Language Model Scaling

Scaling laws predict loss from compute but not how capabilities interact. We measure the coupling between reasoning and truthfulness across 63 base models from 16 families and find a regime change invisible to loss curves: below a family-dependent critical scale N_c, capabilities anticorrelate (r = -0.989, p = 4 x 10^{-5} nonparametric permutation test); above it, they cooperate. N_c ~ 3.5B parameters [2.9B, 13.4B] (bootstrap 95% CI), but model size is not the only variable that determines phase. Architecture, data curation, and training recipe each shift N_c independently: curated training eliminated the coupling dip between Qwen generations (0.025 to 0.830 at matched scale), Gemma-4 at 4B achieves coupling 0.871, characteristic of 13B+ standard-trained models, through distillation and architectural innovation, and Phi at 1B matches web-trained coupling at 10B through data curation alone. Width normalization eliminates the anticorrelation across all tested families, supporting an output-projection bottleneck. Internally, 38 of 40 models show zero competing attention heads. A sparse-regression ODE cross-predicts held-out Llama-2 at 5.6% error. The diagnostic requires no model internals -- only public benchmark scores across a model family. The cooperative regime extends to the frontier (r = +0.72, 34 models, 10 labs). A proof-of-concept intervention confirms the bottleneck is exploitable: adding a single truth-direction vector at the identified layer corrects 60% of misaligned outputs in the tax phase with zero retraining -- a surgical, per-inference correction that requires no weight modification. Code, data, an open-source steering CLI for any open-weight model, and an interactive dashboard for phase diagnosis are released: https://zehenlabs.com/cape/.

cs.LG↗

arXiv2026

The Growing Pains of Frontier Models: When Leaderboards Stop Separating and What to Measure Next

Leaderboards rank frontier models on independent axes but do not reveal whether capabilities reinforce or trade off across releases -- and at the frontier, this interaction is the more informative signal. We decompose paired SWE-bench and GPQA Diamond scores into a population coupling trend and per-release residual ($h$-field) that diagnoses capability emphasis from two public benchmark scores. Across 34 models from 10 labs (2024--2026), capabilities cooperate ($r = +0.72$, $p < 10^{-6}$), but cooperation varies systematically: per-lab coupling slopes span $5\times$ (Google $1.15$ vs. DeepSeek $0.23$), and labs pivot -- DeepSeek reversed from reasoning-rich to coding-first ($\Delta h = 15.9$~pp); Anthropic oscillates between coding excursions and recovery. The population regression serves as an isocline phase boundary: the same $\sqrt{(a/b)\cdot B_1}$ classifier that identifies the base-scale coupling transition [Amin, 2026] classifies frontier models and already detects mixed-phase behavior at the next transition (two models below the GPQA--IFEval isocline). The $h$-field is not just diagnostic -- it tells you what to change. Pretraining establishes coupling at $0.871$ while RLHF adds $0.081$ [Amin, 2026]: pretraining-level shifts are permanent (DeepSeek's four-release reversal persists), post-training shifts are reversible (Anthropic's three coding excursions each recover within one release), and inference compute alone shifts $h$ by $+7.8$~pp without retraining. Knowing which component dominates determines whether to retrain or wait. We provide a three-step diagnostic (locate, classify, predict), a per-lab measurement-priority table, and seven falsifiable predictions with timestamped criteria. Five post-cutoff releases fall within the 95\% prediction interval. Code, data, and an interactive dashboard: https://zehenlabs.com/cape/.

cs.LG↗

arXiv2024

Performance Classification and Remaining Useful Life Prediction of Lithium Batteries Using Machine Learning and Early Cycle Electrochemical Impedance Spectroscopy Measurements

We presents an approach for early cycle classification of lithium-ion batteries into high and low-performing categories, coupled with the prediction of their remaining useful life (RUL) using a linear lasso technique. Traditional methods often rely on extensive cycling and the measurement of a large number of electrochemical impedance spectroscopy (EIS) frequencies to assess battery performance, which can be time and resource consuming. In this study, we propose a methodology that leverages specific EIS frequencies to achieve accurate classification and RUL prediction within the first few cycles of battery operation. Notably, given only the 20 kHz impedance response, our support vector machine (SVM) model classifies batteries with 100\% accuracy. Additionally, our findings reveal that battery performance classification is frequency agnostic within the high frequency ($<20$ kHz) to low-frequency (32 mHz) range. Our model also demonstrates accurate RUL predictions with $R^2>0.96$ based on the out of phase impedance response at a single high (20 kHz) and a single mid-frequency (8.8 Hz), in conjunction with temperature data. This research underscores the significance of the mid-frequency impedance response as merely one among several crucial features in determining battery performance, thereby broadening the understanding of factors influencing battery behavior.

cond-mat.mtrl-sci↗

arXiv2023

Kramers' degenerate magnetism and superconductivity

Motivated by the recent discovery of odd-parity multipolar antiferromagnetic order in CeRh$_2$As$_2$, we examine the coexistence of such translation invariant Kramers' degenerate magnetic states and superconductivity. We show that the presence of such magnetic states generically suppresses superconductivity, whether it be spin-singlet or spin-triplet, unless the magnetic state drives a symmetry-required pair density wave (PDW) superconducting order. We apply our results to CeRh$_2$As$_2$, where no pair density wave order appears; and to the loop current order in the cuprates, where such pair density wave superconductivity must appear together with Bogoliubov Fermi surfaces. In the former case, we explain why superconductivity is not suppressed.

cond-mat.supr-con↗

arXiv2023

Nematic Bogoliubov Fermi surfaces from magnetic toroidal order in FeSe$_{1-x}$S$_x$

Recently it has been argued that the superconducting state of FeSe$_{1-x}$S$_x$ exhibits Bogoliubov Fermi surfaces for $x>0.17$. These Bogoliubov Fermi surfaces appear together with broken time-reversal symmetry and surprisingly demonstrate nematic behavior in a structurally tetragonal phase. Through a symmetry-based analysis of Bogoliubov Fermi surfaces that can arise from broken time-reversal symmetry, we argue that the likely origin of time-reversal symmetry breaking is due to magnetic toroidal order. We show that this magnetic toroidal order naturally appears as a consequence of either static N\'{e}el antiferromagnetic order or due to the formation of a spontaneous pair density wave superconducting order. Finally, we reveal that independent of the presence of Bogoliubov Fermi surfaces, supercurrents will induce N\'{e}el magnetic order in many Fe-based superconductors.

cond-mat.supr-con↗

arXiv2020

Generalized Spin Fluctuation Feedback In Correlated Fermion Superconductors

Experiments reveal that the superconductors $\text{UPt}_3$, $\text{U}_{1-x}\text{Th}_x\text{Be}_{13}$ and $\text{PrOs}_4\text{Sb}_{12}$ undergo two superconducting transitions in the absence of an applied magnetic field. The prevalence of these multiple transitions suggests a common underlying mechanism. A natural candidate theory which accounts for these two transitions is the existence of a small symmetry breaking field, however such a field has not been observed in $\text{PrOs}_4\text{Sb}_{12}$ or $\text{U}_{1-x}\text{Th}_x\text{Be}_{13}$ and has been called into question for $\text{UPt}_3$. Motivated by arguments originally developed for superfluid $^3\text{He}$ we propose that a generalized spin fluctuation feedback effect is responsible for these two transitions. We first develop a phenomenological theory for $^3\text{He}$ that couples spin fluctuations to superfluidity, which correctly predicts that a high temperature broken time-reversal superfluid $^3\text{He}$ phase can emerge as a consequence. The transition at lower temperatures into a time-reversal invariant superfluid phase must then be first order by symmetry arguments. We then apply this phenomenological approach to the three superconductors $\text{UPt}_3$, $\text{U}_{1-x}\text{Th}_x\text{Be}_{13}$ and $\text{PrOs}_4\text{Sb}_{12}$ revealing that this naturally leads to a high-temperature time-reversal invariant nematic superconducting phase, which can be followed by a second order phase transition into a broken time-reversal symmetry phase, as observed.

cond-mat.supr-con↗