arXiv subjects

Tongxi Wang

Publications and source records attributed to Tongxi Wang.

8 recordsLinked to original sources

arXiv2026

Hidden Gauge Controls Feature Specialization in ReLU Networks

The success of deep learning depends on learning useful representations, yet predicting how training organizes these representations across neurons remains difficult. In this work, we show that changing the scale of initial weights can determine which neurons learn a feature without altering any neuron's initial contribution. We construct ReLU networks with identical initial features and predictions that reach the same final predictions with different roles for their neurons. In one, all neurons share the learned feature. In the other, one neuron acquires it while every other neuron's contribution vanishes. The only change is the relative scale of each neuron's input and output weights. Our analysis explains how an initial learning advantage persists through convergence: as one neuron learns the target, it reduces the error driving the others and limits their subsequent adaptation. We prove this outcome in a nonlinear model under gradient flow and small-step gradient descent, and quantify how scale changes the speed and path of feature learning. Experiments verify the predicted dynamics and show that scale also changes feature assignment when two features compete. These results reveal how initialization can control the organization of a learned representation without changing what the network initially represents.

cs.LG↗

arXiv2026

ChiEngMixBench: Evaluating Large Language Models on Expert-Style Chinese-English Terminology Mixing

Large language models increasingly mediate multilingual professional communication, where useful generation requires adapting to community conventions about which expressions are retained, translated, or mixed. Existing benchmarks rarely isolate such community-conditioned choices. We introduce ChiEngMixBench, a controlled benchmark for Chinese AI/CS discourse, where Chinese frames routinely incorporate established English technical terms. Built from public technical discussions, it contains 1,706 source-derived candidate pairs covering 1,344 non-empty normalized terms, including a 1,167-pair strict subset that fixes the Chinese prefix and syntactic position while varying only the terminology form. The benchmark combines paired likelihood comparisons with a transparent reference-profile diagnostic for open-ended responses. Across nine open-weight models, Chinese equivalents receive higher likelihood on most pairs, revealing a gap between source-attested usage and model preference. Specialized terms show a small directional lift that is not robust after frequency and length controls and multiple-comparison correction. Human evaluation and baseline analyses show that reference-profile conformity is informative under the intended mixed-style rubric but does not reliably predict holistic response preference. ChiEngMixBench provides a reusable testbed for community-specific multilingual conventions with explicit diagnostic boundaries.

cs.CL↗

arXiv2026

Tracking Drift: Variation-Aware Entropy Scheduling for Non-Stationary Reinforcement Learning

Real-world reinforcement learning often faces environment drift, but most existing methods rely on static entropy coefficients/target entropy, causing over-exploration during stable periods and under-exploration after drift, and leaving unanswered the principled question of how exploration intensity should scale with drift magnitude. We show that, under standard assumptions, entropy scheduling in non-stationary maximum-entropy RL can be cast as the dynamic-regret trade-off between tracking a drifting comparator and stabilizing updates, yielding a square-root scaling rule for the entropy weight in terms of a online non-stationarity proxy. Building on this, we propose AES--Adaptive Entropy Scheduling--which adaptively adjusts the entropy coefficient/temperature online using observable drift proxies during training, requiring almost no structural changes and incurring minimal overhead. Across 4 algorithm variants, 12 tasks, and 4 drift modes, AES significantly reduces the fraction of performance degradation caused by drift and accelerates recovery after abrupt changes.

cs.LG↗

arXiv2026

Sharp Spectral Thresholds for Logit Fixed Points

Softmax feedback systems are a common mathematical core of entropy-regularized reinforcement learning, logit game dynamics, population choice, and mean-field variational updates. Their central stability question is simple: when does a self-reinforcing softmax system produce a unique and globally predictable outcome? Classical theory gives a conservative answer. By treating softmax as a unit-scale response, it certifies stability only in a strongly randomized regime. We prove that the classical approach misses an entire stable regime and does not identify the point at which the qualitative change truly occurs. For finite-dimensional affine logit systems, the sharp dimension-free Euclidean threshold is $$β\|ΠWΠ\|_{\mathcal T\to\mathcal T}<2,$$ rather than the previously used condition, which certifies stability only while the softmax system remains safely over-regularized. Our theorem fills the previously missing pre-bifurcation regime, extending stability guarantees for affine softmax feedback systems to reward-responsive yet globally predictable systems. It enlarges the certified stability boundary for these systems and identifies where the model genuinely undergoes a phase transition.

cs.LG↗

arXiv2026

FBS: Modeling Native Parallel Reading inside a Transformer

Large language models (LLMs) excel across many tasks, yet inference is still dominated by strictly token-by-token autoregression. Existing acceleration methods largely patch this pipeline and miss core human-reading ingredients: content-adaptive foresight, chunk-structure-aware compute allocation, and train-test consistency for preview/skimming. We propose the Fovea-Block-Skip Transformer (FBS), which injects a causal, trainable loop into Transformers via Parafovea-Attention Window (PAW), Chunk-Head (CH), and Skip-Gate (SG). Across diverse benchmarks, FBS improves the quality-efficiency trade-off without increasing parameters, and ablations show the three modules are complementary.

cs.AI↗

arXiv2026

Stability of In-Context Learning: A Spectral Coverage Perspective

In-context learning (ICL) is a pivotal capability for the practical deployment of large-scale language models, yet its reliability can vary substantially with the number of demonstrations provided in the prompt. A central obstacle is that the target notion, \emph{distributional stability under demonstration resampling}, is expensive to measure directly at scale, making prompt-length selection largely heuristic. We therefore study a \emph{computable sufficient condition} based on a spectral-coverage proxy: the lower tail of the spectrum of a regularized empirical second-moment matrix formed from demonstration representations. Under sub-Gaussian representation assumptions, we derive a non-asymptotic sample-size requirement (a lower bound on $K$) that guarantees this proxy event with prescribed failure probability, yielding a conservative prompt-length recommendation produced by an observable two-stage estimator. In large-scale experiments, the resulting estimates consistently upper-bound empirical accuracy knee-points, which we treat only as a practical surrogate for the prompt-length transition rather than a definition of stability. On a smaller held-out subset, direct resampling-based distributional stability measurements further validate the intended stability interpretation. Finally, a validation-only calibration step tightens the conservatism (typically to about $1.03$--$1.20\times$) while preserving conservative ordering, providing practical and verifiable guidance for ICL prompt design.

cs.LG↗

arXiv2025

Via Score to Performance: Efficient Human-Controllable Long Song Generation with Bar-Level Symbolic Notation

Song generation is regarded as the most challenging problem in music AIGC; nonetheless, existing approaches have yet to fully overcome four persistent limitations: controllability, generalizability, perceptual quality, and duration. We argue that these shortcomings stem primarily from the prevailing paradigm of attempting to learn music theory directly from raw audio, a task that remains prohibitively difficult for current models. To address this, we present Bar-level AI Composing Helper (BACH), the first model explicitly designed for song generation through human-editable symbolic scores. BACH introduces a tokenization strategy and a symbolic generative procedure tailored to hierarchical song structure. Consequently, it achieves substantial gains in the efficiency, duration, and perceptual quality of song generation. Experiments demonstrate that BACH, with a small model size, establishes a new SOTA among all publicly reported song generation systems, even surpassing commercial solutions such as Suno. Human evaluations further confirm its superiority across multiple subjective metrics.

cs.SD↗

arXiv2025

Research on Non-Contact Resistance

This paper investigated the phenomenon of non-contact resistance by inserting a non-magnetic metal rod into an induction coil to explore the response changes of an LRC circuit. We focused on analyzing the changes in inductance when non-ferromagnetic materials (such as H59 brass) were inserted into the coil and verified the impact of the copper rod on inductance through theoretical derivation and experimental validation. Based on Maxwell's equations, the magnetic field distribution within the copper rod was thoroughly derived, and the inductance and resistance values were experimentally measured. These results confirm the accuracy of the theoretical model.

physics.class-ph↗